We precisely characterize the expressivity of computable Recurrent Graph Neural Networks (recurrent GNNs). We prove that recurrent GNNs with finite-precision parameters, sum aggregation, and ReLU activation, can compute any graph algorithm that respects the natural message-passing invariance induced by the Color Refinement (or Weisfeiler-Leman) algorithm. While it is well known that the expressive power of GNNs is limited by this invariance [Morris et al., AAAI 2019; Xu et al., ICLR 2019], we establish that recurrent GNNs can actually match this limit. This is in contrast to non-recurrent GNNs, which have the power of Weisfeiler-Leman only in a very weak, "non-uniform", sense where each graph size requires a different GNN to compute with. Our construction introduces only a polynomial overhead in both time and space. Furthermore, we show that by incorporating random initialization, for connected graphs recurrent GNNs can express all graph algorithms. In particular, any polynomial-time graph algorithm can be emulated on connected graphs in polynomial time by a recurrent GNN with random initialization.

We revisit the algorithmic problem of reconstructing a graph from homomorphism counts that has first been studied in (Böker et al., STACS 2024): given graphs F₁,…,F_k and counts m₁,…,m_k, decide if there is a graph G such that the number of homomorphisms from F_i to G is m_i, for all i. We prove that the problem is NEXP-hard if the counts m_i are specified in binary and Σ₂^p-complete if they are in unary. Furthermore, as a positive result, we show that the unary version can be solved in polynomial time if the constraint graphs are stars of bounded size.

Within the model-theoretic framework for supervised learning introduced by Grohe and Turán (TOCS 2004), we study the parameterized complexity of learning concepts definable in monadic second-order logic (MSO). We show that the problem of learning an MSO-definable concept from a training sequence of labeled examples is fixed-parameter tractable on graphs of bounded clique-width, and that it is hard for the parameterized complexity class para-NP on general graphs. It turns out that an important distinction to be made is between 1-dimensional and higher-dimensional concepts, where the instances of a k-dimensional concept are k-tuples of vertices of a graph. For the higher-dimensional case, we give a learning algorithm that is fixed-parameter tractable in the size of the graph, but not in the size of the training sequence, and we give a hardness result showing that this is optimal. By comparison, in the 1-dimensional case, we obtain an algorithm that is fixed-parameter tractable in both.

We lay the foundations for a database-inspired approach to interpreting and understanding neural network models by querying them using declarative languages. Towards this end we study different query languages, based on first-order logic, that mainly differ in their access to the neural network model. First-order logic over the reals naturally yields a language which views the network as a black box; only the input-output function defined by the network can be queried. This is essentially the approach of constraint query languages. On the other hand, a white-box language can be obtained by viewing the network as a weighted graph, and extending first-order logic with summation over weight terms. The latter approach is essentially an abstraction of SQL. In general, the two approaches are incomparable in expressive power, as we will show. Under natural circumstances, however, the white-box approach can subsume the black-box approach; this is our main result. We prove the result concretely for linear constraint queries over real functions definable by feedforward neural networks with a fixed number of hidden layers and piecewise linear activation functions.

The k -dimensional Weisfeiler-Leman ( k -WL) algorithm is a simple combinatorial algorithm that was originally designed as a graph isomorphism heuristic. It naturally finds applications in Babai’s quasipolynomial-time isomorphism algorithm, practical isomorphism solvers, and algebraic graph theory. However, it also has surprising connections to other areas such as logic, proof complexity, combinatorial optimization, and machine learning. The algorithm iteratively computes a coloring of the k -tuples of vertices of a graph. Since Fürer’s linear lower bound [ICALP 2001], it has been an open question whether there is a super-linear lower bound for the iteration number for k -WL on graphs. We answer this question affirmatively, establishing an Ω ( n k /2 )-lower bound for all k .

We prove new upper and lower bounds on the number of iterations the \(k\) -dimensional Weisfeiler-Leman algorithm ( \(k\) -WL) requires until stabilization. For \(k\geq 3\) , we show that \(k\) -WL stabilizes after at most \(O(kn^{k-1}\log n)\) iterations (where \(n\) denotes the number of vertices of the input structures), obtaining the first improvement over the trivial upper bound of \(n^{k}-1\) and extending a previous upper bound of \(O(n\log n)\) for \(k=2\) . We complement our upper bounds by constructing \(k\) -ary relational structures on which \(k\) -WL requires at least \(n^{\Omega(k)}\) iterations to stabilize. This improves over a previous lower bound of \(n^{\Omega(k/\log k)}\) . We also investigate tradeoffs between the dimension and the iteration number of WL, and show that \(d\) -WL, where \(d=\lceil\frac{3(k + 1)}{2}\rceil\) , can simulate the \(k\) -WL algorithm using only \(O(k^{2}\cdot n^{\lfloor k/2\rfloor+1}\log n)\) many iterations, but still requires at least \(n^{\Omega(k)}\) iterations for any \(d\) (that is sufficiently smaller than \(n\) ). The number of iterations required by \(k\) -WL to distinguish two structures corresponds to the quantifier rank of a sentence distinguishing them in the \((k + 1)\) -variable fragment \(\mathsf{C}_{k + 1}\) of first-order logic with counting quantifiers. Hence, our results also imply new upper and lower bounds on the quantifier rank required in the logic \(\mathsf{C}_{k + 1}\) , as well as tradeoffs between variable number and quantifier rank.

We propose and study a framework for quantifying the importance of the choices of parameter values to the result of a query over a database. These parameters occur as constants in logical queries, such as conjunctive queries. In our framework, the importance of a parameter is its SHAP score - a popular instantiation of the game-theoretic Shapley value to measure the importance of feature values in machine learning models. We make the case for the rationale of using this score by explaining the intuition behind SHAP, and by showing that we arrive at this score in two different, apparently opposing, approaches to quantifying the contribution of a parameter. The application SHAP requires two components in addition to the query and the database: (a) a probability distribution over the combinations of parameter values, and (b) a utility function that measures the similarity between the result for the original parameters and the result for hypothetical parameters. The main question addressed in the paper is the complexity of calculating the SHAP score for different distributions and similarity measures. In particular, we devise polynomial-time algorithms for the case of full acyclic conjunctive queries for certain (natural) similarity functions. We extend our results to conjunctive queries with parameterized filters (e.g., inequalities between variables and parameters). We also illustrate the application of our results to "why-not" explanations (aiming to explain the absence of a query answer), where we consider the task of quantifying the contribution of query components to the elimination of a non-answer in consideration. Finally, we discuss a simple approximation technique for the case of correlated parameters.

Graph Transformers (GTs) such as SAN and GPS are graph processing models that combine Message-Passing GNNs (MPGNNs) with global Self-Attention. They were shown to be universal function approximators, with two reservations: 1. The initial node features must be augmented with certain positional encodings. 2. The approximation is non-uniform: Graphs of different sizes may require a different approximating network. We first clarify that this form of universality is not unique to GTs: Using the same positional encodings, also pure MPGNNs and even 2-layer MLPs are non-uniform universal approximators. We then consider uniform expressivity: The target function is to be approximated by a single network for graphs of all sizes. There, we compare GTs to the more efficient MPGNN + Virtual Node architecture. The essential difference between the two model definitions is in their global computation method: Self-Attention Vs Virtual Node. We prove that none of the models is a uniform-universal approximator, before proving our main result: Neither model’s uniform expressivity subsumes the other’s. We demonstrate the theory with experiments on synthetic data. We further augment our study with real-world datasets, observing mixed results which indicate no clear ranking in practice as well.

Machine learning on graphs, especially using graph neural networks (GNNs), has seen a surge in interest due to the wide availability of graph data across a broad spectrum of disciplines, from life to social and engineering sciences. Despite their practical success, our theoretical understanding of the properties of GNNs remains highly incomplete. Recent theoretical advancements primarily focus on elucidating the coarse-grained expressive power of GNNs, predominantly employing combinatorial techniques. However, these studies do not perfectly align with practice, particularly in understanding the generalization behavior of GNNs when trained with stochastic first-order optimization techniques. In this position paper, we argue that the graph machine learning community needs to shift its attention to developing a balanced theory of graph machine learning, focusing on a more thorough understanding of the interplay of expressive power, generalization, and optimization.

Graph neural networks (GNN) are deep learning architectures for graphs. Essentially, a GNN is a distributed message passing algorithm, which is controlled by parameters learned from data. It operates on the vertices of a graph: in each iteration, vertices receive a message on each incoming edge, aggregate these messages, and then update their state based on their current state and the aggregated messages. The expressivity of GNNs can be characterised in terms of certain fragments of first-order logic with counting and the Weisfeiler-Lehman algorithm.

A common interpretation of soft constraints penalizes the database for every violation of every constraint, where the penalty is the cost (weight) of the constraint. A computational challenge is that of finding an optimal subset: a collection of database tuples that minimizes the total penalty when each tuple has a cost of being excluded. When the constraints are strict (i.e., have an infinite cost), this subset is a “cardinality repair” of an inconsistent database; in soft interpretations, this subset corresponds to a “most probable world” of a probabilistic database, a “most likely intention” of a probabilistic unclean database, and so on. Within the class of functional dependencies, the complexity of finding a cardinality repair is thoroughly understood. Yet, very little is known about the complexity of finding an optimal subset for the more general soft semantics. The work described in this manuscript makes significant progress in that direction. In addition to general insights about the hardness and approximability of the problem, we present algorithms for two special cases (and some generalizations thereof): a single functional dependency, and a bipartite matching. The latter is the problem of finding an optimal “almost matching” of a bipartite graph where a penalty is paid for every lost edge and every violation of monogamy. For these special cases, we also investigate the complexity of additional computational tasks that arise when the soft constraints are used as a means to represent a probabilistic database in the case of a probabilistic unclean database.

Machinery for data analysis often requires a numeric representation of the input. Towards that, a common practice is to embed components of structured data into a high-dimensional vector space. We study the embedding of the tuples of a relational database, where existing techniques are often based on optimization tasks over a collection of random walks from the database. The focus of this paper is on the recent FoRWaRD algorithm that is designed for dynamic databases, where walks are sampled by following foreign keys between tuples. Importantly, different walks have different schemas, or ?walk schemes," that are derived by listing the relations and attributes along the walk. Also importantly, different walk schemes describe relationships of different natures in the database.

The k-dimensional Weisfeiler-Leman (k-WL) algorithm is a simple combinatorial algorithm that was originally designed as a graph isomorphism heuristic. It naturally finds applications in Babai’s quasipolynomial-time isomorphism algorithm, practical isomorphism solvers, and algebraic graph theory. However, it also has surprising connections to other areas such as logic, proof complexity, combinatorial optimization, and machine learning. The algorithm iteratively computes a coloring of the k-tuples of vertices of a graph. Since Fürer’s linear lower bound [ICALP 2001], it has been an open question whether there is a super-linear lower bound for the iteration number for k-WL on graphs. We answer this question affirmatively, establishing an $\Omega\left(n^{k / 2}\right)$-lower bound for all k.

We study the problem of computing an embedding of the tuples of a relational database in a manner that is extensible to dynamic changes of the database. In this problem, the embedding should be stable in the sense that it should not change on the existing tuples due to the embedding of newly inserted tuples (as database applications might already rely on existing embeddings); at the same time, the embedding of all tuples, old and new, should retain high quality. This task is challenging since inter-dependencies among the embeddings of different entities are inherent in state-of-the-art embedding techniques for structured data.We study two approaches to solving the problem. The first is an adaptation of Node2Vec to dynamic databases. The second is the FoRWaRD algorithm (Foreign Key Random Walk Embeddings for Relational Databases) that draws from embedding techniques for general graphs and knowledge graphs, and is inherently utilizing the schema and its key and foreign-key constraints. We evaluate the embedding algorithms using a collection of downstream tasks of column prediction over geographical and biological domains. We find that in the traditional static setting, our two embedding methods achieve comparable results that are compatible with the state-of-the-art for the specific applications. In the dynamic setting, we find that the FoRWaRD algorithm generally outperforms and runs faster than the alternatives, and moreover, it features only a mild reduction of quality even when the database consists of more than half newly inserted tuples after the initial training of the embedding.

We study the complexity of evaluating queries on probabilistic databases under bag semantics. We focus on self-join free conjunctive queries, and probabilistic databases where occurrences of different facts are independent, which is the natural generalization of tuple-independent probabilistic databases to the bag semantics setting. For set semantics, the data complexity of this problem is well understood, even for the more general class of unions of conjunctive queries: it is either in polynomial time, or #P-hard, depending on the query (Dalvi & Suciu, JACM 2012). A reasonably general model of bag probabilistic databases may have unbounded multiplicities. In this case, the probabilistic database is no longer finite, and a careful treatment of representation mechanisms is required. Moreover, the answer to a Boolean query is a probability distribution over (possibly all) non-negative integers, rather than a probability distribution over { true, false }. Therefore, we discuss two flavors of probabilistic query evaluation: computing expectations of answer tuple multiplicities, and computing the probability that a tuple is contained in the answer at most k times for some parameter k. Subject to mild technical assumptions on the representation systems, it turns out that expectations are easy to compute, even for unions of conjunctive queries. For query answer probabilities, we obtain a dichotomy between solvability in polynomial time and #P-hardness for self-join free conjunctive queries.

Recently, many works studied the expressive power of graph neural networks (GNNs) by linking it to the $1$-dimensional Weisfeiler-Leman algorithm ($1\text{-}\mathsf{WL}$). Here, the $1\text{-}\mathsf{WL}$ is a well-studied heuristic for the graph isomorphism problem, which iteratively colors or partitions a graph’s vertex set. While this connection has led to significant advances in understanding and enhancing GNNs’ expressive power, it does not provide insights into their generalization performance, i.e., their ability to make meaningful predictions beyond the training set. In this paper, we study GNNs’ generalization ability through the lens of Vapnik-Chervonenkis (VC) dimension theory in two settings, focusing on graph-level predictions. First, when no upper bound on the graphs’ order is known, we show that the bitlength of GNNs’ weights tightly bounds their VC dimension. Further, we derive an upper bound for GNNs’ VC dimension using the number of colors produced by the $1\text{-}\mathsf{WL}$. Secondly, when an upper bound on the graphs’ order is known, we show a tight connection between the number of graphs distinguishable by the $1\text{-}\mathsf{WL}$ and GNNs’ VC dimension. Our empirical study confirms the validity of our theoretical findings.

The expressivity of Graph Neural Networks (GNNs) is dependent on the aggregation functions they employ. Theoretical works have pointed towards Sum aggregation GNNs subsuming every other GNNs, while certain practical works have observed a clear advantage to using Mean and Max. An examination of the theoretical guarantee identifies two caveats. First, it is size-restricted, that is, the power of every specific GNN is limited to graphs of a specific size. Successfully processing larger graphs may require an other GNN, and so on. Second, it concerns the power to distinguish non-isomorphic graphs, not the power to approximate general functions on graphs, and the former does not necessarily imply the latter. It is desired that a GNN's usability will not be limited to graphs of any specific size. Therefore, we explore the realm of unrestricted-size expressivity. We prove that basic functions, which can be computed exactly by Mean or Max GNNs, are inapproximable by any Sum GNN. We prove that under certain restrictions, every Mean or Max GNN can be approximated by a Sum GNN, but even there, a combination of (Sum, [Mean/Max]) is more expressive than Sum alone. Lastly, we prove further expressivity limitations for GNNs with a broad class of aggregations.

We propose a universal Graph Neural Network architecture which can be trained as an end-2-end search heuristic for any Constraint Satisfaction Problem (CSP). Our architecture can be trained unsupervised with policy gradient descent to generate problem specific heuristics for any CSP in a purely data driven manner. The approach is based on a novel graph representation for CSPs that is both generic and compact and enables us to process every possible CSP instance with one GNN, regardless of constraint arity, relations or domain size. Unlike previous RL-based methods, we operate on a global search action space and allow our GNN to modify any number of variables in every step of the stochastic search. This enables our method to properly leverage the inherent parallelism of GNNs. We perform a thorough empirical evaluation where we learn heuristics for well known and important CSPs, both decision and optimisation problems, from random data, including graph coloring, MAXCUT, and MAX-k-SAT, and the general RB model. Our approach significantly outperforms prior end-2-end approaches for neural combinatorial optimization. It can compete with conventional heuristics and solvers on test instances that are several orders of magnitude larger and structurally more complex than those seen during training.

The fixed-point logic LREC=was developed by Grohe et al. (CSL 2011) in the quest for a logic to capture all problems decidable in logarithmic space. It extends FO+C, first-order logic with counting, by an operator that formalises a limited form of recursion. We show that for every LREC=-definable property on relational structures, there is a constant k such that the k-variable fragment of first-order logic with counting quantifiers expresses the property via formulae of logarithmic quantifier depth. This yields that any pair of graphs separable by the property can be distinguished with the k-dimensional Weisfeiler–Leman algorithm in a logarithmic number of iterations. In particular, it implies that a constant dimension of the algorithm identifies every interval graph and every chordal claw-free graph in logarithmically many iterations, since every such graph admits LREC=-definable canonisation.

We analyse the power of graph neural networks (GNNs) in terms of Boolean circuit complexity and descriptive complexity.We prove that the graph queries that can be computed by a polynomial-size bounded-depth family of GNNs are exactly those definable in the guarded fragment GFO+C of first-order logic with counting and with built-in relations. This puts GNNs in the circuit complexity class TC0. Remarkably, the GNN families may use arbitrary real weights and a wide class of activation functions that includes the standard ReLU, logistic "sigmoid", and hyperbolic tangent functions. If the GNNs are allowed to use random initialisation and global readout (both standard features of GNNs widely used in practice), they can compute exactly the same queries as bounded depth Boolean circuits with threshold gates, that is, exactly the queries in TC0.Moreover, we show that queries computable by a single GNN with piecewise linear activations and rational weights are definable in GFO+C without built-in relations. Therefore, they are contained in uniform TC0.

We prove new upper and lower bounds on the number of iterations the k-dimensional Weisfeiler-Leman algorithm (k-WL) requires until stabilization. For k ≥ 3, we show that k-WL stabilizes after at most O(knk−1log n) iterations (where n denotes the number of vertices of the input structures), obtaining the first improvement over the trivial upper bound of nk− 1 and extending a previous upper bound of O(n log n) for k = 2 [Lichter et al., LICS 2019].We complement our upper bounds by constructing k-ary relational structures on which k-WL requires at least nΩ(k)iterations to stabilize. This improves over a previous lower bound of nΩ(k/logk)[Berkholz, Nordström, LICS 2016].We also investigate tradeoffs between the dimension and the iteration number of WL, and show that d-WL, where $d = \left\lceil {\frac{{3(k + 1)}}{2}} \right\rceil $, can simulate the k-WL algorithm using only O(k2• n⌊k/2⌋+1log n) many iterations, but still requires at least nΩ(k)iterations for any d (that is sufficiently smaller than n).The number of iterations required by k-WL to distinguish two structures corresponds to the quantifier rank of a sentence distinguishing them in the (k + 1)-variable fragment ${{\mathcal{C}}_k}_{ + 1}$ of first-order logic with counting quantifiers. Hence, our results also imply new upper and lower bounds on the quantifier rank required in the logic ${{\mathcal{C}}_k}_{ + 1}$, as well as tradeoffs between variable number and quantifier rank.

In recent years, algorithms and neural architectures based on the Weisfeiler–Leman algorithm, a well-known heuristic for the graph isomorphism problem, have emerged as a powerful tool for machine learning with graphs and relational data. Here, we give a comprehensive overview of the algorithm’s use in a machine-learning setting, focusing on the supervised regime. We discuss the theoretical background, show how to use it for supervised graph and node representation learning, discuss recent extensions, and outline the algorithm’s connection to (permutation-)equivariant neural architectures. Moreover, we give an overview of current applications and future directions to stimulate further research.

In a recent breakthrough, Babai [ Proceedings of STOC, ACM, New York, 2016, pp. 684--697] gave a quasipolynomial-time graph isomorphism test. In this work, we give an improved isomorphism test for graphs of small degree: our algorithm runs in time $n^{\mathcal{O}((\log d)^{c})}$, where $n$ is the number of vertices of the input graphs, $d$ is the maximum degree of the input graphs, and $c$ is an absolute constant. The best previous isomorphism test for graphs of maximum degree $d$ due to Babai, Kantor, and Luks [ Proceedings of FOCS, IEEE, New York, 1983, pp. 162--171] runs in time $n^{\mathcal{O}(d/ \log d)}$.

Abstract. We prove that there is a graph isomorphism test running in time [Formula: see text] on [Formula: see text]-vertex graphs excluding some [Formula: see text]-vertex graph as a minor. Previously known bounds were [Formula: see text] [I. N. Ponomarenko, J. Soviet Math., 55 (1991), pp. 1621–1643] and [Formula: see text] [L. Babai, Proceedings of the 48 th Annual ACM Symposium on Theory of Computing, 2016, pp. 684–697]. For the algorithm we combine recent advances in the group-theoretic graph isomorphism machinery with new graph-theoretic arguments.

We prove that the combinatorial Weisfeiler-Leman algorithm of dimension (3 k +4) is a complete isomorphism test for the class of all graphs of rank width at most k . Rank width is a graph invariant that, similarly to tree width, measures the width of a certain style of hierarchical decomposition of graphs; it is equivalent to clique width. It was known that isomorphism of graphs of rank width k is decidable in polynomial time (Grohe and Schweitzer, FOCS 2015), but the best previously known algorithm has a running time n f(k) for a non-elementary function f . Our result yields an isomorphism test for graphs of rank width k running in time n O(k) . Another consequence of our result is the first polynomial-time canonisation algorithm for graphs of bounded rank width. Our second main result is that fixed-point logic with counting captures polynomial time on all graph classes of bounded rank width.

Lovász (1967) showed that two graphs G and H are isomorphic if and only if they are homomorphism indistinguishable over the class of all graphs, i.e. for every graph F, the number of homomorphisms from F to G equals the number of homomorphisms from F to H. Recently, homomorphism indistinguishability over restricted classes of graphs such as bounded treewidth, bounded treedepth and planar graphs, has emerged as a surprisingly powerful framework for capturing diverse equivalence relations on graphs arising from logical equivalence and algebraic equation systems. In this paper, we provide a unified algebraic framework for such results by examining the linear-algebraic and representation-theoretic structure of tensors counting homomorphisms from labelled graphs. The existence of certain linear transformations between such homomorphism tensor subspaces can be interpreted both as homomorphism indistinguishability over a graph class and as feasibility of an equational system. Following this framework, we obtain characterisations of homomorphism indistinguishability over several natural graph classes, namely trees of bounded degree, graphs of bounded pathwidth (answering a question of Dell et al. (2018)), and graphs of bounded treedepth.

Quantifying the similarity between two graphs is a fundamental algorithmic problem at the heart of many data analysis tasks for graph-based data. In this paper, we study the computational complexity of a family of similarity measures based on quantifying the mismatch between the two graphs, that is, the "symmetric difference" of the graphs under an optimal alignment of the vertices. An important example is similarity based on graph edit distance. While edit distance calculates the "global" mismatch, that is, the number of edges in the symmetric difference, our main focus is on "local" measures calculating the maximum mismatch per vertex. Mathematically, our similarity measures are best expressed in terms of the adjacency matrices: the mismatch between graphs is expressed as the difference of their adjacency matrices (under an optimal alignment), and we measure it by applying some matrix norm. Roughly speaking, global measures like graph edit distance correspond to entrywise matrix norms like the Frobenius norm and local measures correspond to operator norms like the spectral norm. We prove a number of strong NP-hardness and inapproximability results even for very restricted graph classes such as bounded-degree trees.

We analyse the complexity of learning first-order queries in a model-theoretic framework for supervised learning introduced by (Grohe and Turán, TOCS 2004). Previous research on the complexity of learning in this framework focussed on the question of when learning is possible in time sublinear in the background structure.

Arguing for the need to combine declarative and probabilistic programming, Bárány et al. (TODS 2017) recently introduced a probabilistic extension of Datalog as a “purely declarative probabilistic programming language.” We revisit this language and propose a more principled approach towards defining its semantics based on stochastic kernels and Markov processes—standard notions from probability theory. This allows us to extend the semantics to continuous probability distributions, thereby settling an open problem posed by Bárány et al. We show that our semantics is fairly robust, allowing both parallel execution and arbitrary chase orders when evaluating a program. We cast our semantics in the framework of infinite probabilistic databases (Grohe and Lindner, LMCS 2022) and show that the semantics remains meaningful even when the input of a probabilistic Datalog program is an arbitrary probabilistic database.

Probabilistic databases (PDBs) model uncertainty in data. The current standard is to view PDBs as finite probability spaces over relational database instances. Since many attributes in typical databases have infinite domains, such as integers, strings, or real numbers, it is often more natural to view PDBs as infinite probability spaces over database instances. In this article, we lay the mathematical foundations of infinite probabilistic databases. Our focus then is on independence assumptions. Tuple-independent PDBs play a central role in theory and practice of PDBs. Here we study infinite tuple-independent PDBs as well as related models such as infinite block-independent disjoint PDBs. While the standard model of PDBs focuses on a set-based semantics, we also study tuple-independent PDBs with a bag semantics and independence in PDBs over uncountable fact spaces. We also propose a new approach to PDBs with an open-world assumption, addressing issues raised by Ceylan et al. (Proc. KR 2016) and generalizing their work, which is still rooted in finite tuple-independent PDBs. Moreover, for countable PDBs we propose an approximate query answering algorithm.

Probabilistic databases (PDBs) model uncertainty in data in a quantitative way. In the established formal framework, probabilistic (relational) databases are finite probability spaces over relational database instances. This finiteness can clash with intuitive query behavior (Ceylan et al., KR 2016), and with application scenarios that are better modeled by continuous probability distributions (Dalvi et al., CACM 2009). We formally introduced infinite PDBs in (Grohe and Lindner, PODS 2019) with a primary focus on countably infinite spaces. However, an extension beyond countable probability spaces raises nontrivial foundational issues concerned with the measurability of events and queries and ultimately with the question whether queries have a well-defined semantics. We argue that finite point processes are an appropriate model from probability theory for dealing with general probabilistic databases. This allows us to construct suitable (uncountable) probability spaces of database instances in a systematic way. Our main technical results are measurability statements for relational algebra queries as well as aggregate queries and Datalog queries.

The Weisfeiler-Leman (WL) algorithm is a well-known combinatorial procedure for detecting symmetries in graphs and it is widely used in graph-isomorphism tests. It proceeds by iteratively refining a colouring of vertex tuples. The number of iterations needed to obtain the final output is crucial for the parallelisability of the algorithm. We show that there is a constant k such that every planar graph can be identified (that is, distinguished from every non-isomorphic graph) by the k-dimensional WL algorithm within a logarithmic number of iterations. This generalises a result due to Verbitsky (STACS 2007), who proved the same for 3-connected planar graphs. The number of iterations needed by the k-dimensional WL algorithm to identify a graph corresponds to the quantifier depth of a sentence that defines the graph in the (k+1)-variable fragment C^{k+1} of first-order logic with counting quantifiers. Thus, our result implies that every planar graph is definable with a C^{k+1}-sentence of logarithmic quantifier depth.

A common interpretation of soft constraints penalizes the database for every violation of every constraint, where the penalty is the cost (weight) of the constraint. A computational challenge is that of finding an optimal subset: a collection of database tuples that minimizes the total penalty when each tuple has a cost of being excluded. When the constraints are strict (i.e., have an infinite cost), this subset is a "cardinality repair" of an inconsistent database; in soft interpretations, this subset corresponds to a "most probable world" of a probabilistic database, a "most likely intention" of a probabilistic unclean database, and so on. Within the class of functional dependencies, the complexity of finding a cardinality repair is thoroughly understood. Yet, very little is known about the complexity of finding an optimal subset for the more general soft semantics. This paper makes a significant progress in this direction. In addition to general insights about the hardness and approximability of the problem, we present algorithms for two special cases: a single functional dependency, and a bipartite matching. The latter is the problem of finding an optimal "almost matching" of a bipartite graph where a penalty is paid for every lost edge and every violation of monogamy.

Graph neural networks (GNNs) are effective models for representation learning on relational data. However, standard GNNs are limited in their expressive power, as they cannot distinguish graphs beyond the capability of the Weisfeiler-Leman graph isomorphism heuristic. In order to break this expressiveness barrier, GNNs have been enhanced with random node initialization (RNI), where the idea is to train and run the models with randomized initial node features. In this work, we analyze the expressive power of GNNs with RNI, and prove that these models are universal, a first such result for GNNs not relying on computationally demanding higher-order properties. This universality result holds even with partially randomized initial node features, and preserves the invariance properties of GNNs in expectation. We then empirically analyze the effect of RNI on GNNs, based on carefully constructed datasets. Our empirical findings support the superior performance of GNNs with RNI over standard GNNs.

Graph neural networks (GNNs) are deep learning architectures for machine learning problems on graphs. It has recently been shown that the expressiveness of GNNs can be characterised precisely by the combinatorial Weisfeiler-Leman algorithms and by finite variable counting logics. The correspondence has even led to new, higher-order GNNs corresponding to the WL algorithm in higher dimensions.The purpose of this paper is to explain these descriptive characterisations of GNNs.

Infinite probabilistic databases (PDBs) are a framework of probability distributions over infinitely many different database instances. We systematically study the representability problem for such PDBs by means of tuple-independence and first-order views. Although first-order views over tuple-independent PDBs are not a complete representation system for infinite PDBs, they form a fairly robust class: Adding first-order constraints does not give them additional expressive power, and they cover many relevant special cases such as block-independent disjoint PDBs, and PDBs of bounded instance size. We identify criteria for representability (or non-representability) in this class and explore their limits. In addition, we study the expressive power of fragments of first-order logic over tuple-independent PDBs. In general, for infinite PDBs, the landscape of relative expressive power for various classes of views over PDBs with independence assumptions is much more complex than in the finite setting.

We introduce the framework of Deep Weisfeiler Leman algorithms (DeepWL), which allows the design of purely combinatorial graph isomorphism tests that are more powerful than the well-known Weisfeiler-Leman algorithm. We prove that, as an abstract computational model, polynomial-time DeepWL-algorithms have exactly the same expressiveness as the logic Choiceless Polynomial Time (with counting) introduced by Blass, Gurevich, and Shelah (Ann. Pure Appl. Logic., 1999). It is a well-known open question whether the existence of a polynomial-time graph isomorphism test implies the existence of a polynomial-time canonisation algorithm. Our main technical result states that for each class of graphs (satisfying some mild closure condition), if there is a polynomial-time DeepWL isomorphism test, then there is a polynomial-time canonisation algorithm for this class. This implies that there is also a logic capturing polynomial time on this class.

We prove that there is a graph isomorphism test running in time npolylog(h)on n-vertex graphs excluding some h-vertex graph as a minor. Previously known bounds were npoly(h)(Ponomarenko, 1988) and npolylog(n)(Babai, STOC 2016). For the algorithm we combine recent advances in the group-theoretic graph isomorphism machinery with new graph-theoretic arguments.

We prove that graphs G, G' satisfy the same sentences of first-order logic with counting of quantifier rank at most k if and only if they are homomorphism-indistinguishable over the class of all graphs of tree depth at most k. Here G, G' are homomorphism-indistinguishable over a class F of graphs if for each graph F ϵ F, the number of homomorphisms from F to G equals the number of homomorphisms from F to G'.

Vector representations of graphs and relational structures, whether hand-crafted feature vectors or learned representations, enable us to apply standard data analysis and machine learning techniques to the structures. A wide range of methods for generating such embeddings have been studied in the machine learning and knowledge representation literature. However, vector embeddings have received relatively little attention from a theoretical point of view. Starting with a survey of embedding techniques that have been used in practice, in this paper we propose two theoretical approaches that we see as central for understanding the foundations of vector embeddings. We draw connections between the various approaches and suggest directions for future research.

Arguing for the need to combine declarative and probabilistic programming, Bárány et al. (TODS 2017) recently introduced a probabilistic extension of Datalog as a "purely declarative probabilistic programming language." We revisit this language and propose a more foundational approach towards defining its semantics. It is based on standard notions from probability theory known as stochastic kernels and Markov processes. This allows us to extend the semantics to continuous probability distributions, thereby settling an open problem posed by Bárány et al. We show that our semantics is fairly robust, allowing both parallel execution and arbitrary chase orders when evaluating a program. We cast our semantics in the framework of infinite probabilistic databases (Grohe and Lindner, ICDT 2020), and we show that the semantics remains meaningful even when the input of a probabilistic Datalog program is an arbitrary probabilistic database.

The Weisfeiler-Leman algorithm is a well-known combinatorial graph isomorphism test going back to work of Weisfeiler and Leman in the late 1960s. The algorithm has a surprising number of seemingly unrelated characterisations in terms of logic, algebra, linear and semi-definite programming, and graph homomorphisms. Due to its simplicity and efficiency, it is an important subroutine of all modern graph isomorphism tools. In recent years, further applications in linear optimisation, probabilistic inference, and machine learning have surfaced. In the first part of my talk, I will give an introduction to the Weisfeiler-Leman algorithm and its various characterisations. In the second part I will speak about its applications, in particular about recent work relating the algorithm to graph neural networks.

We give a new FPT algorithm testing isomorphism of n -vertex graphs of tree-width k in time 2 kpolylog(k) n 3 , improving the FPT algorithm due to Lokshtanov, Pilipczuk, Pilipczuk, and Saurabh (FOCS 2014), which runs in time 2 O(k5 log k) n 5 . Based on an improved version of the isomorphism-invariant graph decomposition technique introduced by Lokshtanov et al., we prove restrictions on the structure of the automorphism groups of graphs of tree-width k . Our algorithm then makes heavy use of the group theoretic techniques introduced by Luks (JCSS 1982) in his isomorphism test for bounded degree graphs and Babai (STOC 2016) in his quasipolynomial isomorphism test. In fact, we even use Babai’s algorithm as a black box in one place. We also give a second algorithm that, at the price of a slightly worse running time 2 O(k2 log k) n 3 , avoids the use of Babai’s algorithm and, more importantly, has the additional benefit that it can also be used as a canonization algorithm.

In recent years, graph neural networks (GNNs) have emerged as a powerful neural architecture to learn vector representations of nodes and graphs in a supervised, end-to-end fashion. Up to now, GNNs have only been evaluated empirically—showing promising results. The following work investigates GNNs from a theoretical point of view and relates them to the 1-dimensional Weisfeiler-Leman graph isomorphism heuristic (1-WL). We show that GNNs have the same expressiveness as the 1-WL in terms of distinguishing non-isomorphic (sub-)graphs. Hence, both algorithms also have the same shortcomings. Based on this, we propose a generalization of GNNs, so-called k-dimensional GNNs (k-GNNs), which can take higher-order graph structures at multiple scales into account. These higher-order structures play an essential role in the characterization of social networks and molecule graphs. Our experimental evaluation confirms our theoretical findings as well as confirms that higher-order information is useful in the task of graph classification and regression.

The Weisfeiler-Leman (WL) dimension of a graph is a measure for the inherent descriptive complexity of the graph. While originally derived from a combinatorial graph isomorphism test called the Weisfeiler-Leman algorithm, the WL dimension can also be characterised in terms of the number of variables that is required to describe the graph up to isomorphism in first-order logic with counting quantifiers. It is known that the WL dimension is upper-bounded for all graphs that exclude some fixed graph as a minor [M. Grohe, 2017]. However, the bounds that can be derived from this general result are astronomic. Only recently, it was proved that the WL dimension of planar graphs is at most 3 [S. Kiefer et al., 2017]. In this paper, we prove that the WL dimension of graphs embeddable in a surface of Euler genus g is at most 4g+3. For the WL dimension of graphs embeddable in an orientable surface of Euler genus g, our approach yields an upper bound of 2g + 3.

We prove that the combinatorial Weisfeiler-Leman algorithm of dimension (3k+4) is a complete isomorphism test for the class of all graphs of rank width at most k. Rank width is a graph invariant that, similarly to tree width, measures the width of a certain style of hierarchical decomposition of graphs; it is equivalent to clique width. It was known that isomorphism of graphs of rank width k is decidable in polynomial time (Grohe and Schweitzer, FOCS 2015), but the best previously known algorithm has a running time nf(k)for a non-elementary function f. Our result yields an isomorphism test for graphs of rank width k running in time nO(k). Another consequence of our result is the first polynomial time canonisation algorithm for graphs of bounded rank width. Our second main result is that fixed-point logic with counting captures polynomial time on all graph classes of bounded rank width.

For every graph class {F}, let HomInd({F}) be the problem of deciding whether two given graphs are homomorphism-indistinguishable over {F}, i.e., for every graph F in {F}, the number hom(F, G) of homomorphisms from F to G equals the corresponding number hom(F, H) for H. For several natural graph classes (such as paths, trees, bounded treewidth graphs), homomorphism-indistinguishability over the class has an efficient structural characterization, resulting in polynomial time solvability [H. Dell et al., 2018]. In particular, it is known that two non-isomorphic graphs are homomorphism-indistinguishable over the class {T}_k of graphs of treewidth k if and only if they are not distinguished by k-dimensional Weisfeiler-Leman algorithm, a central heuristic for isomorphism testing: this characterization implies a polynomial time algorithm for HomInd({T}_k), for every fixed k in N. In this paper, we show that there is a polynomial-time-decidable class {F} of undirected graphs of bounded treewidth such that HomInd({F}) is undecidable. Our second hardness result concerns the class {K} of complete graphs. We show that HomInd({K}) is co-NP-hard, and in fact, we show completeness for the class C_=P (under P-time Turing reductions). On the algorithmic side, we show that HomInd({P}) can be solved in polynomial time for the class {P} of directed paths. We end with a brief study of two variants of the HomInd({F}) problem: (a) the problem of lexographic-comparison of homomorphism numbers of two graphs, and (b) the problem of computing certain distance-measures (defined via homomorphism numbers) between two graphs.

Probabilistic databases (PDBs) introduce uncertainty into relational databases by specifying probabilities for several possible instances. Traditionally, they are finite probability spaces over database instances. Such finite PDBs inherently make a closed-world assumption: non-occurring facts are assumed to be impossible, rather than just unlikely. As convincingly argued by Ceylan et al. (KR '16), this results in implausibilities and clashes with intuition. An open-world assumption, where facts not explicitly listed may have a small positive probability can yield more reasonable results. The corresponding open-world model of Ceylan et al., however, assumes that all entities in the PDB come from a fixed finite universe. In this work, we take one further step and propose a model of "truly" open-world PDBs with an infinite universe. This is natural when we consider entities from typical domains such as integers, real numbers, or strings. While the probability space might become infinitely large, all instances of a PDB remain finite. We provide a sound mathematical framework for infinite PDBs generalizing the existing theory of finite PDBs. Our main results are concerned with countable, tuple-independent PDBs; we present a generic construction showing that such PDBs exist in the infinite and provide a characterization of their existence. This construction can be used to give an open-world semantics to finite PDBs. The construction can also be extended to so-called block-independent-disjoint probabilistic databases. Algorithmic questions are not the focus of this paper, but we show how query evaluation algorithms can be lifted from finite PDBs to perform approximate evaluation (with an arbitrarily small additive approximation error) in countably infinite tuple-independent PDBs.

We establish new, and surprisingly tight, connections between propositional proof complexity and finite model theory. Specifically, we show that the power of several propositional proof systems, such as Horn resolution, bounded-width resolution, and the monomial calculus of bounded degree, can be characterised in a precise sense by variants of fixed-point logics that are of fundamental importance in descriptive complexity theory. Our main results are that Horn resolution has the same expressive power as least fixed-point logic, that bounded-width resolution captures existential least fixed-point logic, and that the polynomial calculus with bounded degree over the rationals solves precisely the problems definable in fixed-point logic with counting. We also study the bounded-degree polynomial calculus. Over the rationals, it captures fixed-point logic with counting if we restrict the bit-complexity of the coefficients. For unrestricted coefficients, we can only say that the bounded-degree polynomial calculus is at most as powerful as bounded variable infinitary counting logic, but a precise logical characterisation of its power remains an open problem. These connections between logics and proof systems allow us to establish finite-model-theoretic tools for proving lower bounds for the polynomial calculus over the rationals and also over finite fields. This is a corrected version of the paper (arXiv:1802.09377) published originally on January 23, 2019.

In a recent breakthrough, Babai (STOC 2016) gave quasipolynomial graph isomorphism test. In this work, we give an improved isomorphism test for graphs of small degree: our algorithms runs in time n^O((log d)^c), where n is the number of vertices of the input graphs, d is the maximum degree of the input graphs, and c is an absolute constant. The best previous isomorphism test for graphs of maximum degree d due to Babai, Kantor and Luks (FOCS 1983) runs in time n^O(d log d).

In this paper, we relate a beautiful theory by Lovász with a popular heuristic algorithm for the graph isomorphism problem, namely the color refinement algorithm and its k-dimensional generalization known as the Weisfeiler-Leman algorithm. We prove that two graphs G and H are indistinguishable by the color refinement algorithm if and only if, for all trees T, the number Hom(T,G) of homomorphisms from T to G equals the corresponding number Hom(T,H) for H. There is a natural system of linear equations whose nonnegative integer solutions correspond to the isomorphisms between two graphs. The nonnegative real solutions to this system are called fractional isomorphisms, and two graphs are fractionally isomorphic if and only if the color refinement algorithm cannot distinguish them (Tinhofer 1986, 1991). We show that, if we drop the nonnegativity constraints, that is, if we look for arbitrary real solutions, then a solution to the linear system exists if and only if, for all t, the two graphs have the same number of length-t walks. We lift the results for trees to an equivalence between numbers of homomorphisms from graphs of tree width k, the k-dimensional Weisfeiler-Leman algorithm, and the level-k Sherali-Adams relaxation of our linear program. We also obtain a partial result for graphs of bounded path width and solutions to our system where we drop the nonnegativity constraints. A consequence of our results is a quasi-linear time algorithm to decide whether, for two given graphs G and H, there is a tree T with Hom(T,G) = Hom(T,H).

We prove that for every positive integer k, there exists an MSO1-transduction that given a graph of linear cliquewidth at most k outputs, nondeterministically, some clique decomposition of the graph of width bounded by a function of k. A direct corollary of this result is the equivalence of the notions of CMSO1-definability and recognizability on graphs of bounded linear cliquewidth.

The graph similarity problem, also known as approximate graph isomorphism or graph matching problem, has been extensively studied in the machine learning community, but has not received much attention in the algorithms community: Given two graphs G,H of the same order n with adjacency matrices A_G,A_H, a well-studied measure of similarity is the Frobenius distance dist(G,H):=min_{pi}|A_G^{pi}-A_H|_F, where pi ranges over all permutations of the vertex set of G, where A_G^pi denotes the matrix obtained from A_G by permuting rows and columns according to pi, and where |M |_F is the Frobenius norm of a matrix M. The (weighted) graph similarity problem, denoted by GSim (WSim), is the problem of computing this distance for two graphs of same order. This problem is closely related to the notoriously hard quadratic assignment problem (QAP), which is known to be NP-hard even for severely restricted cases. It is known that GSim (WSim) is NP-hard; we strengthen this hardness result by showing that the problem remains NP-hard even for the class of trees. Identifying the boundary of tractability for WSim is best done in the framework of linear algebra. We show that WSim is NP-hard as long as one of the matrices has unbounded rank or negative eigenvalues: hence, the realm of tractability is restricted to positive semi-definite matrices of bounded rank. Our main result is a polynomial time algorithm for the special case where the associated (weighted) adjacency graph for one of the matrices has a bounded number of twin equivalence classes. The key parameter underlying our algorithm is the clustering number of a graph; this parameter arises in context of the spectral graph drawing machinery.

We study an extension of first-order logic FO that allows to express cardinality conditions in a similar way as SQL's COUNT operator. The corresponding logic FOC(P) was introduced by Kuske and Schweikardt, who showed that query evaluation for this logic is fixed-parameter tractable on classes of databases of bounded degree. In this paper, we first show that the fixed-parameter tractability of FOC(P) cannot even be generalised to very simple classes of databases of unbounded degree such as unranked trees or strings with a linear order relation. Then, we identify a fragment FOC1(P) of FOCP which is still extends FO and is sufficiently strong to express standard applications of SQL's COUNT operator. Our main result shows that query evaluation for FOC1(P) is fixed-parameter tractable on nowhere dense classes of databases. This, in particular, implies that the counting problem for first-order queries on nowhere dense classes is fixed-parameter tractable.

In [M. Grohe, S. Kreutzer, and S. Siebertz, J. ACM, 64 (2017), 17] it was shown that nowhere dense classes of graphs admit sparse neighborhood covers of small degree. We show that a monotone graph class admits sparse neighborhood covers if and only if it is nowhere dense. The existence of such covers for nowhere dense classes is established through bounds on so-called weak coloring numbers. The core results of this paper are various lower and upper bounds on the weak coloring numbers and other, closely related, generalized coloring numbers. We prove tight bounds for these numbers on graphs of bounded treewidth. We clarify and tighten the relation between the density of shallow minors and the various generalized coloring numbers. These upper bounds are complemented by new, stronger exponential lower bounds on the weak and strong coloring numbers, and by superpolynomial lower bounds on the weak coloring numbers on classes of polynomial expansion. Finally, we show that computing weak $r$-coloring numbers is NP-complete for all $r\geq 3$.

We study the classification problems over string data for hypotheses specified by formulas of monadic second-order logic MSO. The goal is to design learning algorithms that run in time polynomial in the size of the training set, independently of or at least sublinear in the size of the whole data set. We prove negative as well as positive results. If the data set is an unprocessed string to which our algorithms have local access, then learning in sublinear time is impossible even for hypotheses definable in a small fragment of first-order logic. If we allow for a linear time pre-processing of the string data to build an index data structure, then learning of MSO-definable hypotheses is possible in time polynomial in the size of the training set, independently of the size of the whole data set.

We prove that the solvability of systems of linear equations and related linear algebraic properties are definable in a fragment of fixed-point logic with counting that only allows polylogarithmically many iterations of the fixed-point operators. This enables us to separate the descriptive complexity of solving linear equations from full fixed-point logic with counting by logical means. As an application of these results, we separate an extension of first-order logic with a rank operator from fixed-point logic with counting, solving an open problem due to Holm [21]. We then draw a connection from this work in descriptive complexity theory to graph isomorphism testing and propositional proof complexity. Answering an open question from [7], we separate the strength of certain algebraic graph-isomorphism tests. This result can also be phrased as a separation of the algebraic propositional proof systems “Nullstellensatz” and “monomial PC”.

We consider a declarative framework for machine learning where concepts and hypotheses are defined by formulas of a logic over some “background structure”. We show that within this framework, concepts defined by first-order formulas over a background structure of at most polylogarithmic degree can be learned in polylogarithmic time in the “probably approximately correct” learning sense.

In recent years, we have seen several approaches to the graph isomorphism problem based on “generic” mathematical programming or algebraic (Grobner basis) techniques. For most of these, lower bounds have been established. In fact, it has been shown that the pairs of non-isomorphic CFI-graphs (introduced by Cai, FUrer, and Immerman in 1992 as hard examples for the combinatorial Weisfeiler-Leman algorithm) cannot be distinguished by these mathematical algorithms. A notable exception were the algebraic algorithms over the field 2, for which no lower bound was known. Another, in some way even stronger, approach to graph isomorphism testing is based on solving systems of linear Diophantine equations (that is, linear equations over the integers), which is known to be possible in polynomial time. So far, no lower bounds for this approach were known. Lower bounds for the algebraic algorithms can best be proved in the framework of proof complexity, where they can be phrased as lower bounds for algebraic proof systems such as Nullstellensatz or the (more powerful) polynomial calculus. We give new hard examples for these systems: families of pairs of non-isomorphic graphs that are hard to distinguish by polynomial calculus proofs simultaneously over all prime fields, including 2, as well as examples that are hard to distinguish by the systems-of-linear-Diophantine- equations approach. In a previous paper, we observed that the CFI-graphs are closely related to what we call “group CSPs”: constraint satisfaction problems where the constraints are membership tests in some coset of a subgroup of a cartesian power of a base group (ℤ2 in the case of the classical CFI-graphs). Our new examples are also based on group CSPs (for Abelian groups), but here we extend the CSPs by a few non-group constraints to obtain even harder instances for graph isomorphism.

Nowhere dense graph classes, introduced by Nešetřil and Ossona de Mendez [2010, 2011], form a large variety of classes of “sparse graphs” including the class of planar graphs, actually all classes with excluded minors, and also bounded degree graphs and graph classes of bounded expansion. We show that deciding properties of graphs definable in first-order logic is fixed-parameter tractable on nowhere dense graph classes (parameterized by the length of the input formula). At least for graph classes closed under taking subgraphs, this result is optimal: it was known before that for all classes C of graphs closed under taking subgraphs, if deciding first-order properties of graphs in C is fixed-parameter tractable, then C must be nowhere dense (under a reasonable complexity theoretic assumption). As a by-product, we give an algorithmic construction of sparse neighborhood covers for nowhere dense graphs. This extends and improves previous constructions of neighborhood covers for graph classes with excluded minors. At the same time, our construction is considerably simpler than those. Our proofs are based on a new game-theoretic characterization of nowhere dense graphs that allows for a recursive version of locality-based algorithms on these classes. On the logical side, we prove a “rank-preserving” version of Gaifman’s locality theorem.

An assignment of colours to the vertices of a graph is stable if any two vertices of the same colour have identically coloured neighbourhoods. The goal of colour refinement is to find a stable colouring that uses a minimum number of colours. This is a widely used subroutine for graph isomorphism testing algorithms, since any automorphism needs to be colour preserving. We give an O (( m + n )log n ) algorithm for finding a canonical version of such a stable colouring, on graphs with n vertices and m edges. We show that no faster algorithm is possible, under some modest assumptions about the type of algorithm, which captures all known colour refinement algorithms.

We introduce a new decomposition of a graphs into quasi-4-connected components, where we call a graph quasi-4-connected if it is 3-connected and it only has separations of order 3 that separate a single vertex from the rest of the graph. Moreover, we give a cubic time algorithm computing the decomposition of a given graph. Our decomposition into quasi-4-connected components refines the well-known decompositions of graphs into biconnected and triconnected components. We relate our decomposition to Robertson and Seymour's theory of tangles by establishing a correspondence between the quasi-4-connected components of a graph and its tangles of order 4.

Order-invariant formulas access an ordering on a structure's universe, but the model relation is independent of the used ordering. They are frequently used for logic-based approaches in computer science. Order-invariant formulas capture unordered problems of complexity classes and they model the independence of the answer to a database query from low-level aspects of databases. We study the expressive power of order-invariant monadic second-order (MSO) and first-order (FO) logic on restricted classes of structures that admit certain forms of tree decompositions (not necessarily of bounded width).

Tangles of graphs have been introduced by Robertson and Seymour in the context of their graph minor theory. Tangles may be viewed as describing “$k$-connected components” of a graph (though in a twisted way). An interesting aspect of tangles is that they can be defined not only for graphs, but more generally for arbitrary connectivity functions (that is, integer-valued submodular and symmetric set functions). However, tangles are difficult to deal with algorithmically. To start with, it is unclear how to represent them, because they are families of separations and, as such, may be exponentially large. Our first contribution is a data structure for representing and accessing all tangles of a graph up to some fixed order. Using this data structure, we can prove an algorithmic version of a very general structure theorem due to Carmesin et al. (for graphs) and Hundertmark (for arbitrary connectivity functions) that yields a canonical tree decomposition whose parts correspond to the maximal tangles. This may be viewed as a generalization of the decomposition of a graph into its 3-connected components.

We study on which classes of graphs first-order logic ( fo ) and monadic second-order logic ( mso ) have the same expressive power. We show that for all classes C of graphs that are closed under taking subgraphs, fo and mso have the same expressive power on C if and only if, C has bounded tree depth. Tree depth is a graph invariant that measures the similarity of a graph to a star in a similar way that tree width measures the similarity of a graph to a tree. For classes just closed under taking induced subgraphs, we show an analogous result for guarded second-order logic ( gso ), the variant of mso that not only allows quantification over vertex sets but also over edge sets. A key tool in our proof is a Feferman--Vaught-type theorem that works for infinite collections of structures despite being constructive.

We give an algorithm that, for every fixed k, decides isomorphism of graphs of rank width at most k in polynomial time. As the rank width of a graph is bounded in terms of its clique width, we also obtain a polynomial time isomorphism test for graph classes of bounded clique width.

Tangles of graphs have been introduced by Robertson and Seymour in the context of their graph minor theory. Tangles may be viewed as describing "k-connected components" of a graph (though in a twisted way). They play an important role in graph minor theory. An interesting aspect of tangles is that they cannot only be defined for graphs, but more generally for arbitrary connectivity functions (that is, integer-valued submodular and symmetric set functions).

Abstract We give a new, simplified and detailed account of the correspondence between levels of the Sherali–Adams relaxation of graph isomorphism and levels of pebble-game equivalence with counting (higher-dimensional Weisfeiler–Lehman colour refinement). The correspondence between basic colour refinement and fractional isomorphism, due to Tinhofer [22; 23] and Ramana, Scheinerman and Ullman [17], is re-interpreted as the base level of Sherali–Adams and generalised to higher levels in this sense by Atserias and Maneva [1] and Malkin [14], who prove that the two resulting hierarchies interleave. In carrying this analysis further, we here give (a) a precise characterisation of the level k Sherali–Adams relaxation in terms of a modified counting pebble game; (b) a variant of the Sherali–Adams levels that precisely match the k-pebble counting game; (c) a proof that the interleaving between these two hierarchies is strict. We also investigate the variation based on boolean arithmetic instead of real/rational arithmetic and obtain analogous correspondences and separations for plain k-pebble equivalence (without counting). Our results are driven by considerably simplified accounts of the underlying combinatorics and linear algebra.

We generalize the structure theorem of Robertson and Seymour for graphs excluding a fixed graph $H$ as a minor to graphs excluding $H$ as a topological subgraph. We prove that for a fixed $H$, every graph excluding $H$ as a topological subgraph has a tree decomposition where each part is either “almost embeddable” to a fixed surface or has bounded degree with the exception of a bounded number of vertices. Furthermore, we prove that such a decomposition is computable by an algorithm that is fixed-parameter tractable with parameter $|H|$. We present two algorithmic applications of our structure theorem. To illustrate the mechanics of a “typical” application of the structure theorem, we show that on graphs excluding $H$ as a topological subgraph, Partial Dominating Set (find $k$ vertices whose closed neighborhood has maximum size) can be solved in time $f(H,k)\cdot n^{O(1)}$. More significantly, we show that on graphs excluding $H$ as a topological subgraph, Graph Isomorphism can be solved in time $n^{f(H)}$. This result unifies and generalizes two previously known important polynomial time solvable cases of Graph Isomorphism: bounded-degree graphs [E. M. Luks, J. Comput. System Sci., 25 (1982), pp. 42--65] and $H$-minor free graphs [I. N. Ponomarenko, Zap. Nauchn. Sem. Leningrad. Otdel. Mat. Inst. Steklov. (LOMI), 174 (1988), pp. 147--177, 182]. The proof of this result needs a generalization of our structure theorem to the context of invariant treelike decomposition.

Color refinement is a basic algorithmic routine for graph isomorphismtesting and has recently been used for computing graph kernels as well as for lifting belief propagation and linear programming. So far, color refinement has been treated as a combinatorial problem. Instead, we treat it as a nonlinear continuous optimization problem and prove thatit implements a conditional gradient optimizer that can be turned into graph clustering approaches using hashing and truncated power iterations. This shows that color refinement is easy to understand in terms of random walks, easy to implement (matrix-matrix/vector multiplications) and readily parallelizable. We support our theoretical results with experiments on real-world graphs with millions of edges.

Colour refinement is a simple algorithm that partitions the vertices of a graph according their "iterated degree sequence." It has very efficient implementations, running in quasilinear time, and a surprisingly wide range of applications. The algorithm has been designed in the context of graph isomorphism testing, and it is used an important subroutine in almost all practical graph isomorphism tools. Somewhat surprisingly, other applications in machine learning, probabilistic inference, and linear programming have surfaced recently. In the first part of my talk, I will introduce the basic algorithm as well as higher dimensional extensions known as the k-dimensional Weisfeiler-Lehman algorithm. I will also discuss an unexpected connection between colour refinement and a natural linear programming approach to graph isomorphism testing. In the second part of my talk, I will discuss various applications of colour refinement.

Nowhere dense graph classes, introduced by Nešetřil and Ossona de Mendez [30], form a large variety of classes of "sparse graphs" including the class of planar graphs, actually all classes with excluded minors, and also bounded degree graphs and graph classes of bounded expansion.

Many important combinatorial problems can be modeled as constraint satisfaction problems. Hence, identifying polynomial-time solvable classes of constraint satisfaction problems has received a lot of attention. In this article, we are interested in structural properties that can make the problem tractable. So far, the largest structural class that is known to be polynomial-time solvable is the class of bounded hypertree width instances introduced by Gottlob et al. [2002]. Here we identify a new class of polynomial-time solvable instances: those having bounded fractional edge cover number. Combining hypertree width and fractional edge cover number, we then introduce the notion of fractional hypertree width. We prove that constraint satisfaction problems with bounded fractional hypertree width can be solved in polynomial time (provided that the tree decomposition is given in the input). Together with a recent approximation algorithm for finding such decompositions [Marx 2010], it follows that bounded fractional hypertree width is now the most generally known structural property that guarantees polynomial-time solvability.

Nowhere dense classes of graphs were introduced by Nesetril and Ossona de Mendez as a model for "sparsity" in graphs. It turns out that nowhere dense classes of graphs can be characterised in many different ways and have been shown to be equivalent to other concepts studied in areas such as (finite) model theory. Therefore, the concept of nowhere density seems to capture a natural property of graph classes generalising for example classes of graphs which exclude a fixed minor, have bounded degree or bounded local tree-width. In this paper we give a self-contained introduction to the concept of nowhere dense classes of graphs focussing on the various ways in which they can be characterised. We also briefly sketch algorithmic applications these characterisations have found in the literature.

A key result of Robertson and Seymour's graph minor theory is a structure theorem stating that all graphs excluding some fixed graph as a minor have a tree decomposition into pieces that are almost embeddable in a fixed surface. Most algorithmic applications of graph minor theory rely on an algorithmic version of this result. However, the known algorithms for computing such graph minor decompositions heavily rely on the very long and complicated proofs of the existence of such decompositions, essentially they retrace these proofs and show that all steps are algorithmic. In this paper, we give a simple quadratic time algorithm for computing graph minor decompositions. The best previously known algorithm due to Kawarabayashi and Wollan runs in cubic time and is far more complicated. Our algorithm combines techniques from logic and structural graph theory, or more precisely, a variant of Courcelle's Theorem stating that monadic second-order logic formulas can be evaluated in linear time on graphs of bounded tree width and Robertson and Seymour's so called Weak Structure Theorem.

Relational joins are at the core of relational algebra, which in turn is the core of the standard database query language SQL. As their evaluation is expensive and very often dominated by the output size, it is an important task for database query optimizers to compute estimates on the size of joins and to find good execution plans for sequences of joins. We study these problems from a theoretical perspective, both in the worst-case model and in an average-case model where the database is chosen according to a known probability distribution. In the former case, our first key observation is that the worst-case size of a query is characterized by the fractional edge cover number of its underlying hypergraph, a combinatorial parameter previously known to provide an upper bound. We complete the picture by proving a matching lower bound and by showing that there exist queries for which the join-project plan suggested by the fractional edge cover approach may be substantially better than any join plan that does not use intermediate projections. On the other hand, we show that in the average-case model, every join-project plan can be turned into a plan containing no projections in such a way that the expected time to evaluate the plan increases only by a constant factor independent of the size of the database. Not surprisingly, the key combinatorial parameter in this context is the maximum density of the underlying hypergraph. We show how to make effective use of this parameter to eliminate the projections.

We study on which classes of graphs first-order logic (FO) and monadic second-order logic (MSO) have the same expressive power. We show that for each class of graphs that is closed under taking subgraphs, FO and MSO have the same expressive power on the class if, and only if, it has bounded tree depth. Tree depth is a graph invariant that measures the similarity of a graph to a star in a similar way that tree width measures the similarity of a graph to a tree. For classes just closed under taking induced subgraphs, we show an analogous result for guarded second-order logic (GSO), the variant of MSO that not only allows quantification over vertex sets but also over edge sets. A key tool in our proof is a Feferman-Vaught-type theorem that is constructive and still works for unbounded partitions.

The question of whether there is a polynomial time algorithm deciding whether two graphs are isomorphic has been a one of the best known open problems in theoretical computer science for more than forty years. Indeed, the graph isomorphism problem is one of the very few natural problems in NP that is neither known to be in P nor known to be NP-complete. The question is still wide open, but a number of deep partial results giving polynomial time algorithms for specific classes of graphs are known. Many of them have been obtained through a group theoretic approach that dominated the research on the graph isomorphism problem since the early 1980s. After an introductory survey on the graph isomorphism problem, in my talk I will focus on approaches based on structural graph theory and connections between logical definability and certain combinatorial algorithms for the isomorphism problem. In particular, I will speak about two recent results: The first says that the Weisfeiler-Lehman algorithm (a simple combinatorial algorithm) can be used to decide isomorphism on graph classes with excluded minors in polynomial time. The second says that isomorphism can be decided in polynomial time on graph classes with excluded topological subgraphs.

We generalize the structure theorem of Robertson and Seymour for graphs excluding a fixed graph H as a minor to graphs excluding H as a topological subgraph. We prove that for a fixed H, every graph excluding H as a topological subgraph has a tree decomposition where each part is either "almost embeddable" to a fixed surface or has bounded degree with the exception of a bounded number of vertices. Furthermore, such a decomposition is computable by an algorithm that is fixed-parameter tractable with parameter |H|.

We give a logical characterization of the polynomial-time properties of graphs embeddable in some surface. For every surfaceS, a property P of graphs embeddable inSis decidable in polynomial time if and only if it is definable in fixed-point logic with counting. It is a consequence of this result that for every surfaceSthere is aksuch that a simple combinatorial algorithm, namely “thek-dimensional Weisfeiler-Lehman algorithm”, decides isomorphism of graphs embeddable inSin polynomial time. We also present (without proof) generalizations of these results to arbitrary classes of graphs with excluded minors.

The homomorphism problem for relational structures is an abstract way of formulating constraint satisfaction problems (CSP) and various problems in database theory. The decision version of the homomorphism problem received a lot of attention in literature; in particular, the way the graph-theoretical structure of the variables and constraints influences the complexity of the problem is intensively studied. Here we study the problem of enumerating all the solutions with polynomial delay from a similar point of view. It turns out that the enumeration problem behaves very differently from the decision version. We give evidence that it is unlikely that a characterization result similar to the decision version can be obtained. Nevertheless, we show nontrivial cases where enumeration can be done with polynomial delay.

We prove that for every fixed undirected graph H, there is an O(|V(G)|3) time algorithm that, given a graph G, tests if G contains H as a topological subgraph (that is, a subdivision of H is subgraph of G). This shows that topological subgraph testing is fixed-parameter tractable, resolving a longstanding open question of Downey and Fellows from 1992. As a corollary, for every H we obtain an O(|V(G)|3) time algorithm that tests if there is an immersion of H into a given graph G. This answers another open question raised by Downey and Fellows in 1992.

In a fundamental article on query languages for relational databases, Chandra and Harel [2] asked in 1982 whether there is a language that expresses precisely those queries which can be answered in polynomial time. Gurevich [10] later rephrased the question in the language of finite model theory, asking whether there is a logic that captures polynomial time. Despite serious efforts in the late 1980s and the 1990s, the question is still wide open. It is considered to be one of the main open problems in database theory and finite model theory. Recently, there has been a renewed interest in the question. New languages have been proposed [1, 4, 5] and old ones reconsidered [3, 12], and a number of partial results stating that certain languages capture polynomial time on large and natural classes of structures have been obtained [6, 8, 9, 11].

We prove that fixed-point logic with counting captures polynomial time on all classes of graphs with excluded minors. That is, for every class C of graphs such that some graph H is not a minor of any graph in C, a property P of graphs in C is decidable in polynomial time if and only if it is definable in fixed-point logic with counting. Furthermore, we prove that for every class C of graphs with excluded minors there is a k such that the k-dimensional Weisfeiler-Leman algorithm decides isomorphism of graphs in C in polynomial time. The Weisfeiler-Leman algorithm is a combinatorial algorithm for testing isomorphism. It generalises the basic colour refinement algorithm and is much simpler than the known group-theoretic algorithms for deciding isomorphism of graphs with excluded minors. The main technical theorem behind these two results is a "definables tructure theorem" for classes of graphs with excluded minors. It states that graphs with excluded minors can be decomposed into pieces arranged in a treelike structure, together with a linear order of each of the pieces. Furthermore, the decomposition and the linear orders on the pieces are definable in fixed-point logic (without counting).

Partition functions, also known as homomorphism functions, form a rich family of graph invariants that contain combinatorial invariants such as the number of k-colorings or the number of independent sets of a graph and also the partition functions of certain “spin glass” models of statistical physics such as the Ising model. Building on earlier work by Dyer and Greenhill [Random Structures Algorithms, 17 (2000), pp. 260–289] and Bulatov and Grohe [Theoret. Comput. Sci., 348 (2005), pp. 148–186], we completely classify the computational complexity of partition functions. Our main result is a dichotomy theorem stating that every partition function is either computable in polynomial time or #P-complete. Partition functions are described by symmetric matrices with real entries, and we prove that it is decidable in polynomial time in terms of the matrix whether a given partition function is in polynomial time or #P-complete. While in general it is very complicated to give an explicit algebraic or combinatorial description of the tractable cases, for partition functions described by Hadamard matrices (these turn out to be central in our proofs) we obtain a simple algebraic tractability criterion, which says that the tractable cases are those “representable” by a quadratic polynomial over the field $\mathbb{F}_2$.

We introduce extensions of first-order logic (FO) and fixed-point logic (FP) with operators that compute the rank of a definable matrix. These operators are generalizations of the counting operations in FP+C (i.e. fixed-point logic with counting) that allow us to count the dimension of a definable vector space, rather than just count the cardinality of a definable set. The logics we define have data complexity contained in polynomial time and all known examples of polynomial time queries that are not definable in FP+C are definable in FP+rk, the extension of FP with rank operators. For each prime number p and each positive integer n, we have rank operators rkpfor determining the rank of a matrix over the finite field GFpdefined by a formula over n-tuples. We compare the expressive power of the logics obtained by varying the values p and n can take. In particular, we show that increasing the arity of the operators yields an infinite hierarchy of expressive power. The rank operators are surprisingly expressive, even in the absence of fixed-point operators. We show that FO+rkpcan define deterministic and symmetric transitive closure. This allows us to show that, on ordered structures, FO+rkpcaptures the complexity class MODpL, for all prime values of p.

\emph{Partition functions}, also known as \emph{homomorphism functions}, form a rich family of graph invariants that contain combinatorial invariants such as the number of $k$-colourings or the number of independent sets of a graph and also the partition functions of certain ``spin glass'' models of statistical physics such as the Ising model. Building on earlier work by Dyer and Greenhill (2000) and Bulatov and Grohe (2005), we completely classify the computational complexity of partition functions. Our main result is a dichotomy theorem stating that every partition function is either computable in polynomial time or \#P-complete. Partition functions are described by symmetric matrices with real entries, and we prove that it is decidable in polynomial time in terms of the matrix whether a given partition function is in polynomial time or \#P-complete. While in general it is very complicated to give an explicit algebraic or combinatorial description of the tractable cases, for partition functions described by a Hadamard matrices --- these turn out to be central in our proofs --- we obtain a simple algebraic tractability criterion, which says that the tractable cases are those ``representable'' by a quadratic polynomial over the field $\ensuremath{\mathbb{F}_2}$.

We consider a scenario where we want to query a large dataset that is stored in external memory and does not fit into main memory. The most constrained resources in such a situation are the size of the main memory and the number of random accesses to external memory. We note that sequentially streaming data from external memory through main memory is much less prohibitive. We propose an abstract model of this scenario in which we restrict the size of the main memory and the number of random accesses to external memory, but admit arbitrary sequential access. A distinguishing feature of our model is that it allows the usage of unlimited external memory for storing intermediate results, such as several hard disks that can be accessed in parallel. In this model, we prove lower bounds for the problem of sorting a sequence of strings (or numbers), the problem of deciding whether two given sets of strings are equal, and two closely related decision problems. Intuitively, our results say that there is no algorithm for the problems that uses internal memory space bounded by N 1−ε and at most o (log N ) random accesses to external memory, but unlimited “streaming access”, both for writing to and reading from external memory. (Here, N denotes the size of the input and ε is an arbitrary constant greater than 0.) We even permit randomized algorithms with one-sided bounded error. We also consider the problem of evaluating database queries and prove similar lower bounds for evaluating relational algebra queries against relational databases and XQuery and XPath queries against XML-databases.

We prove that the weighted monotone circuit satisfiability problem has no fixed-parameter tractable approximation algorithm with constant or polylogarithmic approximation ratio unless FPT = W[P]. Our result answers a question of Alekhnovich and Razborov, who proved that the weighted monotone circuit satisfiability problem has no fixed-parameter tractable 2-approximation algorithm unless every problem in W[P] can be solved by a randomized fpt algorithm and asked whether their result can be derandomized. Alekhnovich and Razborov used their inapproximability result as a lemma for proving that resolution is not automatizable unless W[P] is contained in randomized FPT. It is an immediate consequence of our result that the complexity theoretic assumption can be weakened to W[P] =!= FPT. The decision version of the monotone circuit satisfiability problem is known to be complete for the class W[P]. By reducing them to the monotone circuit satisfiability problem with suitable approximation preserving reductions, we prove similar inapproximability results for all other natural minimisation problems known to be W[P]-complete.

Relational joins are at the core of relational algebra, which in turn is the core of the standard database query language SQL. As their evaluation is expensive and very often dominated by the output size, it is an important task for database query optimisers to compute estimates on the size of joins and to find good execution plans for sequences of joins. We study these problems from a theoretical perspective, both in the worst-case model, and in an average-case model where the database is chosen according to a known probability distribution. In the former case, our first key observation is that the worst-case size of a query is characterised by the fractional edge cover number of its underlying hypergraph, a combinatorial parameter previously known to provide an upper bound. We complete the picture by proving a matching lower bound, and by showing that there exist queries for which the join-project plan suggested by the fractional edge cover approach may be substantially better than any join plan that does not use intermediate projections.

The question of whether there is a logic that captures polynomial time is the central open problem in descriptive complexity theory. In my talk, I will review the question and the early, mostly negative results that were obtained until the mid 1990s, and then move on to positive results about capturing polynomial time on specific classes of graphs. This will include recent results on definability in fixed-point logic and graph structure theory. Finally, I will dicuss stronger logics and propose directions for further research.The purpose of this accompanying note is to give the basic definitions in detail, state the main results, mention some open problems, and give a list of references.

We introduce a notion of definable tree decompositions of graphs. Actually, a definable tree decomposition of a graph is not just a tree decomposition, but a more complicated structure that represents many different tree decompositions of the graph. It is definable in the graph by a tuple of formulas of some logic. In this paper, only study tree decomposition definable in fixed-point logic. We say that a definable tree decomposition is over a class of graphs if the pieces of the decomposition are in this class. We prove two general theorems lifting definability results from the pieces of a tree decomposition of a graph to the whole graph. Besides unifying earlier work on fixed-point definability and descriptive complexity theory on planar graphs and graphs of bounded tree width, these general results can be used to prove that the class of all graphs without a K5-minor is definable infixed-point logic and that fixed-point logic with counting captures polynomial time on this class.

A class of relational structures is said to have the extension preservation property if every first-order sentence that is preserved under extensions on the class is equivalent to an existential sentence. The class of all finite structures does not have the extension preservation property. We study the property on classes of finite structures that are better behaved. We show that the property holds for classes of acyclic structures, structures of bounded degree, and more generally structures that are wide in a sense that we will make precise. We also show that the preservation property holds for the class of structures of treewidth at most k, for any k. In contrast, we show that the property fails for the class of planar graphs.

We introduce the concept of locally excluded minors. Graph classes locally excluding a minor are a common generalisation of the concept of excluded minor classes and of graph classes with bounded local tree-width. We show that first-order model-checking is fixed-parameter tractable on any class of graphs locally excluding a minor. This strictly generalises analogous results by Flum and Grohe on excluded minor classes and Frick and Grohe on classes with bounded local tree-width. As an important consequence of the proof we obtain fixed-parameter algorithms for problems such as dominating or independent set on graph classes excluding a minor, where now the parameter is the size of the dominating set and the excluded minor. We also study graph classes with excluded minors, where the minor may grow slowly with the size of the graphs and show that again, first-order model-checking is fixed-parameter tractable on any such class of graphs.

We give a complexity theoretic classification of homomorphism problems for graphs and, more generally, relational structures obtained by restricting the left hand side structure in a homomorphism. For every class C of structures, let HOM(C,−) be the problem of deciding whether a given structure A ∈C has a homomorphism to a given (arbitrary) structure ß. We prove that, under some complexity theoretic assumption from parameterized complexity theory, HOM(C,−) is in polynomial time if and only if C has bounded tree width modulo homomorphic equivalence. Translated into the language of constraint satisfaction problems, our result yields a characterization of the tractable structural restrictions of constraint satisfaction problems. Translated into the language of database theory, it implies a characterization of the tractable instances of the evaluation problem for conjunctive queries over relational databases.

Abstract We observe that the W*-hierarchy, a variant (introduced by Downey, Fellows, and Taylor [7]) of the better known W-hierarchy, coincides with the W-hierarchy, though not level wise, but just as a whole hierarchy. More precisely, we prove that W[t] ⊆ W* [t] ⊆ W[2t − 2] for each t ≥ 2. It was known before that W[1] = W*[1] and W[2] = W*[2]. Our second main result is a new logical characterization of the W*-hierarchy in terms of “Fagin-definable problems.” As a by-product, we also obtain an improvement of our earlier characterization of the hierarchy in terms of model-checking problems. Furthermore, we obtain new complete problems for the classes W[3] and W*[3].

We establish a close connection between (sub)exponential time complexity and parameterized complexity by proving that the so-called miniaturization mapping is a reduction preserving isomorphism between the two theories.

We establish a close connection between (sub)exponential time complexity and parameterized complexity by proving that the so-called miniaturization mapping is a reduction preserving isomorphism between the two theories.

Let phi(X) be a first-order formula in the language of graphs that has a free set variable X, and assume that X only occurs positively in phi(X). Then a natural minimisation problem associated with phi(X) is to find, in a given graph G, a vertex set S of minimum size such that G satisfies phi(S). Similarly, if X only occurs negatively in phi(X), then phi(X) defines a maximisation problem. Many well-known optimisation problems are first-order definable in this sense, for example, minimum dominating set or maximum independent set. We prove that for each class Gscr of graphs with excluded minors, in particular for each class of planar graphs, the restriction of a first-order definable optimisation problem to the class Gscr has a polynomial time approximation scheme. A crucial building block of the proof of this approximability result is a version of Gaifman's locality theorem for formulas positive in a set variable. This result may be of independent interest

We study the randomized version of a computation model (introduced in [9, 10]) that restricts random access to external memory and internal memory space. Essentially, this model can be viewed as a powerful version of a data stream model that puts no cost on sequential scans of external memory (as other models for data streams) and, in addition, (like other external memory models, but unlike streaming models), admits several large external memory devices that can be read and written to in parallel.We obtain tight lower bounds for the decision problems set equality, multiset equality, and checksort. More precisely, we show that any randomized one-sided-error bounded Monte Carlo algorithm for these problems must perform Ω(logN) random accesses to external memory devices, provided that the internal memory size is at most O(4√N/logN), where N denotes the size of the input data.From the lower bound on the set equality problem we can infer lower bounds on the worst case data complexity of query evaluation for the languages XQuery, XPath, and relational algebra on streaming data. More precisely, we show that there exist queries in XQuery, XPath, and relational algebra, such that any (randomized) Las Vegas algorithm that evaluates these queries must perform Ω(logN) random accesses to external memory devices, provided that the internal memory size is at most O(4√N/logN).

We consider a scenario where we want to query a large dataset that is stored in external memory and does not fit into main memory. The most constrained resources in such a situation are the size of the main memory and the number of random accesses to external memory. We note that sequentially streaming data from external memory through main memory is much less prohibitive.We propose an abstract model of this scenario in which we restrict the size of the main memory and the number of random accesses to external memory, but do not restrict sequential reads. A distinguishing feature of our model is that it admits the usage of unlimited external memory for storing intermediate results, such as several hard disks that can be accessed in parallel. In practice, such auxiliary external memory can be crucial. For example, in a first sequential pass the data can be annotated, and in a second pass this annotation can be used to answer the query. Koch's [9] ARB system for answering XPath queries is based on such a strategy.In this model, we prove lower bounds for sorting the input data. As opposed to related results for models without auxiliary external memory for intermediate results, we cannot rely on communication complexity to establish these lower bounds. Instead, we simulate. our model by a non-uniform computation model for which we can establish the lower bounds by combinatorial means.

Most parameterized complexity classes are defined in terms of a parameterized version of the Boolean satisfiability problem (the so-called weighted satisfiability problem). For example, Downey and Fellow's W-hierarchy is of this form. But there are also classes, for example, the A-hierarchy, that are more naturally characterised in terms of model-checking problems for certain fragments of first-order logic. Downey, Fellows, and Regan were the first to establish a connection between the two formalisms by giving a characterisation of the W-hierarchy in terms of first-order model-checking problems. We improve their result and then prove a similar correspondence between weighted satisfiability and model-checking problems for the A-hierarchy and the W^*-hierarchy. Thus we obtain very uniform characterisations of many of the most important parameterized complexity classes in both formalisms. Our results can be used to give new, simple proofs of some of the core results of structural parameterized complexity theory.

Succinctness is a natural measure for comparing the strength of different logics. Intuitively, a logic L_1 is more succinct than another logic L_2 if all properties that can be expressed in L_2 can be expressed in L_1 by formulas of (approximately) the same size, but some properties can be expressed in L_1 by (significantly) smaller formulas. We study the succinctness of logics on linear orders. Our first theorem is concerned with the finite variable fragments of first-order logic. We prove that: (i) Up to a polynomial factor, the 2- and the 3-variable fragments of first-order logic on linear orders have the same succinctness. (ii) The 4-variable fragment is exponentially more succinct than the 3-variable fragment. Our second main result compares the succinctness of first-order logic on linear orders with that of monadic second-order logic. We prove that the fragment of monadic second-order logic that has the same expressiveness as first-order logic on linear orders is non-elementarily more succinct than first-order logic.

Most parameterized complexity classes are defined in terms of a parameterized version of the Boolean satisfiability problem (the so-called weighted satisfiability problem. For example, Downey and Fellow's W-hierarchy is of this form. But there are also classes, for example, the A-hierarchy, that are more naturally characterised in terms of model-checking problems for fragments of first-order logic. R. G. Downey et al. (1998) were the first to establish a connection between the two formalisms by giving a characterisation of the W-hierarchy in terms of first-order model-checking problems. We improve their result and then prove a similar correspondence between weighted satisfiability and model-checking problems for the A-hierarchy and the W-hierarchy. Thus we obtain very uniform characterisations of many of the most important parameterized complexity classes in both formalisms. Our results can be used to give new, simple proofs of some of the core results of structural parameterized complexity theory.

Succinctness is a natural measure for comparing the strength of different logics. Intuitively, a logic L/sub 1/ is more succinct than another logic L/sub 2/ if oil properties that can be expressed in L/sub 2/ can be expressed in L/sub 1/ by formulas of (approximately) the same size, but some properties can be expressed in L/sub 1/ by (significantly) smaller formulas. We study the succinctness of logics on linear orders that have the same expressive power as first-order logic. Our first theorem is concerned with the finite variable fragments of first-order logic. We prove that:(i) Up to a polynomial factor, the 2- and the 3-variable fragments of first-order logic on linear orders have the same succinctness.(ii) The 4-variable fragment is exponentially more succinct than the 3-variable fragment. Our second main result compares the succinctness of first-order logic on linear orders with that of monadic second-order logic. We prove that the fragment of monadic second-order logic that has the same expressiveness as first-order logic on linear orders is non-elementarily more succinct than first-order logic.

We develop a parameterized complexity theory for counting problems. As the basis of this theory, we introduce a hierarchy of parameterized counting complexity classes #W$[t]$, for $t\ge 1$, that corresponds to Downey and Fellows's W-hierarchy [R. G. Downey and M. R. Fellows, Parameterized Complexity, Springer-Verlag, New York, 1999] and we show that a few central W-completeness results for decision problems translate to #W-completeness results for the corresponding counting problems. Counting complexity gets interesting with problems whose decision version is tractable, but whose counting version is hard. Our main result states that counting cycles and paths of length k in both directed and undirected graphs, parameterized by k, is #W$[1]-complete. This makes it highly unlikely that these problems are fixed-parameter tractable, even though their decision versions are fixed-parameter tractable. More explicitly, our result shows that most likely there is no $f(k) \cdot n^c$-algorithm for counting cycles or paths of length k in a graph of size n for any computable function $f: \mathbb{N} \to \mathbb{N}$ and constant c, even though there is a $2^{O(k)} \cdot n^{2.376}$ algorithm for finding a cycle or path of length k [N. Alon, R. Yuster, and U. Zwick, J. ACM, 42 (1995), pp. 844--856].

We give machine characterisations and logical descriptions of a number of parameterized complexity classes. The focus of our attention is the class W[P], which we characterise as the class of all parameterized problems decidable by a nondeterministic fixed-parameter tractable algorithm, whose use of nondeterminism is bounded in terms of the parameter. We give similar characterisations for AW[P], the "alternating version of W[P]", and various other parameterized complexity classes. We also give logical characterisations of the classes W[P] and AW[P] in terms of fragments of least fixed-point logic, thereby putting these two classes into a uniform framework that we have developed in earlier work. Furthermore, we investigate the relation between alternation and space in parameterized complexity theory. We prove that the compact Turing machine computation problem, shown to be hard for the class AW[SAT] in (K. A. Abrahamson et al., 1995) is complete for the class uniform-XNL.

We give a complexity theoretic classification of homomorphism problems for graphs and, more generally, relational structures obtained by restricting the left hand side structure in a homomorphism. For every class C of structures, let HOM(C, /spl I.bar/) be the problem of deciding whether a given structure A /spl isin/ C has a homomorphism to a given (arbitrary) structure B. We prove that, under some complexity theoretic assumption from parameterized complexity theory, HOM(C, /spl I.bar/) is in polynomial time if, and only if, the cores of all structures in C have bounded tree-width (as long as the structures in C only contain relations of bounded arity). Due to a well known correspondence between homomorphism problems and constraint satisfaction problems, our classification carries over to the latter.

This paper studies the problem of evaluating unary (or node-selecting) queries on unranked trees compressed in a natural structure-preserving way, by the sharing of common subtrees. The motivation to study unary queries on unranked trees comes from the database field, where querying XML (Extensible Markup Language) documents, which can be considered as unranked labeled trees, is an important task. We give algorithms and complexity results for the evaluation of XPath and monadic datalog queries. Furthermore, we propose a new automata-theoretic formalism for querying trees and give algorithms for evaluating queries defined by such automata.

We develop a parameterized complexity theory for counting problems. As the basis of this theory, we introduce a hierarchy of parameterized counting complexity classes #W[t], for t/spl ges/1, that corresponds to Downey and Fellows' (1999) W-hierarchy and show that a few central W-completeness results for decision problems translate to #W-completeness results for the corresponding counting problems. Counting complexity gets interesting with problems whose decision version is tractable, but whose counting version is hard. Our main result states that counting cycles and paths of length k in both directed and undirected graphs, parameterized by k, are #W[1]-complete. This makes it highly unlikely that any of these problems is fixed-parameter tractable, even though their decision versions are. More explicitly, our result shows that most likely there is no f(k)/spl middot/n/sup c/-algorithm for counting cycles or paths of length k in a graph of size n for any computable function f:/spl Nopf//spl rarr//spl Nopf/ and constant c, even though there is a 2/sup O(k)//spl middot/n/sup 2.376/ algorithm for finding a cycle or path of length k (2).

The model-checking problem for a logic L on a class C of structures asks whether a given L-sentence holds in a given structure in C. In this paper, we give super-exponential lower bounds for fixed-parameter tractable model-checking problems for first-order and monadic second-order logic. We show that unless PTIME=NP, the model-checking problem for monadic second-order logic on finite words is not solvable in time f(k)/spl middot/p(n), for any elementary function f and any polynomial p. Here k denotes the size of the input sentence and n the size of the input word. We prove the same result for first-order logic under a stronger complexity theoretic assumption from parameterized complexity theory. Furthermore, we prove that the model-checking problems for first-order logic on structures of degree 2 and of bounded degree d/spl ges/3 are not solvable in time 2(2/sup o(k)/)/spl middot/p(n) (for degree 2) and 2(2/sup 2o(k)/)/spl middot/p(n) (for degree d), for any polynomial p, again under an assumption from parameterized complexity theory. We match these lower bounds by corresponding upper bounds.

A number of efficient methods for evaluating first-order and monadic-second order queries on finite relational structures are based on tree-decompositions of structures or queries. We systematically study these methods.In the first part of the article, we consider arbitrary formulas on tree-like structures. We generalize a theorem of Courcelle [1990] by showing that on structures of bounded tree-width a monadic second-order formula (with free first- and second-order variables) can be evaluated in time linear in the structure size plus the size of the output.In the second part, we study tree-like formulas on arbitrary structures. We generalize the notions of acyclicity and bounded tree-width from conjunctive queries to arbitrary first-order formulas in a straightforward way and analyze the complexity of evaluating formulas of these fragments. Moreover, we show that the acyclic and bounded tree-width fragments have the same expressive power as the well-known guarded fragment and the finite-variable fragments of first-order logic, respectively.

One important class of spatial database queries is the class of topological queries , that is, queries invariant under homeomorphisms. We study topological queries expressible in the standard query language on spatial databases, first-order logic with various amounts of arithmetic. Our main technical result is a combinatorial characterization of the expressive power of topological first-order logic on regular spatial databases.

This paper gives a short introduction into parameterized complexity theory, aimed towards database theorists interested in this area. The main results presented here classify the evaluation of first-order queries and conjunctive queries as hard parameterized problems.

We show that for every fixed k\ge 0 there is a quadratic time algorithm that decides whether a given graph has crossing number at most k and, if this is the case, computes a drawing of the graph in the plane with at most k crossings.

The evaluation of conjunctive queries is hard both with respect to its combined complexity (NP-complete) and its parameterized complexity (W[1]-complete). It becomes tractable (PTIME for combined complexity, FPT for parameterized complexity), when the underlying graphs of the conjunctive queries have bounded tree-width [2]. We show that, in some sense, this is optimal both with respect to combined and parameterized complexity: For every class C of graphs, the evaluation of all conjunctive queries whose underlying graph is in C is tractable if, and only if, C has bounded tree-width.

We introduce the concept of a class of graphs, or more generally, relational structures, being locally tree-decomposable . There are numerous examples of locally tree-decomposable classes, among them the class of planar graphs and all classes of bounded valence or of bounded tree-width. We also consider a slightly more general concept of a class of structures having bounded local tree-width .We show that for each property ϕ of structures that is definable in first-order logic and for each locally tree-decomposable class C of structures, there is a linear time algorithm deciding whether a given structure A ∈ C has property ϕ. For classes C of bounded local tree-width, we show that for every k ≥ 1 there is an algorithm solving the same problem in time O ( n 1+(1/ k ) ) (where n is the cardinality of the input structure).

In this article, we study parameterized complexity theory from the perspective of logic, or more specifically, descriptive complexity theory. We propose to consider parameterized model-checking problems for various fragments of first-order logic as generic parameterized problems and show how this approach can be useful in studying both fixed-parameter tractability and intractability. For example, we establish the equivalence between the model-checking for existential first-order logic, the homomorphism problem for relational structures, and the substructure isomorphism problem. Our main tractability result shows that model-checking for first-order formulas is fixed-parameter tractable when restricted to a class of input structures with an excluded minor. On the intractability side, for every $t\ge 0$ we prove an equivalence between model-checking for first-order formulas with t quantifier alternations and the parameterized halting problem for alternating Turing machines with t alternations. We discuss the close connection between this alternation hierarchy and Downey and Fellows' W-hierarchy. On a more abstract level, we consider two forms of definability, called Fagin definability and slicewise definability, that are appropriate for describing parameterized problems. We give a characterization of the class FPT of all fixed-parameter tractable problems in terms of slicewise definability in finite variable least fixed-point logic, which is reminiscent of the Immerman--Vardi theorem characterizing the class PTIME in terms of definability in least fixed-point logic.

One important class of spatial database queries is the class of topological queries, i.e. queries invariant under homeomorphisms. We study topological queries expressible in the standard query language on spatial databases, first-order logic with various amounts of arithmetic. Our main technical result is a combinatorial characterization of the expressive power of topological first-order logic on regular spatial databases.

It is known that standard query languages for constraint databases lack the power to express connectivity properties. Such properties are important in the context of geographical databases, where one naturally wishes to ask queries about connectivity (what are the connected components of a given set?) or reachability (is there a path from A to B that lies entirely in a given region?). No existing constraint query languages that allow closed form evaluation can express these properties.

The k-dimensional Weisfeiler-Leman algorithm, for k 1, is a natural and simple combinatorial algorithm attempting to decide whether two given graphs are isomorphic. In this paper, we show that for every surface S (orientable or non-orientable) there is a k 1 such that the k-dimensional WL-algorithm succeeds to decide isomorphism of graphs embeddable in S. To prove this, we use a close connection between the WL-algorithm and denability in certain nite variable logics that has been established by Cai, Furer, and Immerman [7]. 1. INTRODUCTION The graph isomorphism problem asks whether two given graphs are isomorphic. While complexity theoretic results indicate that the isomorphism problem is not NP-complete (if it was, the polynomial hierarchy would collapse to its second level [6; 29]), no polynomial time algorithm for the general problem is known. However, there is a number of important classes of graphs on which the isomorphism problem is known to be solvable in polynomial t...

One of the fundamental results of descriptive complexity theory, due to Immerman [13] and Vardi [18], says that a class of ordered finite structures is definable in fixed-point logic if, and only if, it is computable in polynomial time. Much effort has been spent on the problem of capturing polynomial time, that is, describing all polynomial time computable classes of not necessarily ordered finite structures by a logic in a similar way. The most obvious shortcoming of fixed-point logic itself on unordered structures is that it cannot count. Immerman [14] responded to this by adding counting constructs to fixed-point logic. Although it has been proved by Cai, Fürer, and Immerman [1] that the resulting fixed-point logic with counting, denoted by IFP+C, still does not capture all of polynomial time, it does capture polynomial time on several important classes of structures (on trees, planar graphs, structures of bounded tree-width [15, 9, 10]). The main motivation for such capturing results is that they may give a better understanding of polynomial time. But of course this requires that the logical side is well understood. We hope that our analysis of IFP+C-formulas will help to clarify the expressive power of IFP+C; in particular, we derive a normal form. Moreover, we obtain a problem complete for IFP+C under first-order reductions.

A query is local if the decision of whether a tuple in a structure satisfies this query only depends on a small neighborhood of the tuple. We prove that all queries expressible by order-invariant first-order formulas are local.

We study the expressive power of inflationary fixed-point logic IFP and inflationary fixed-point logic with counting IFP+C on planar graphs. We prove the following results: (1) IFP captures polynomial time on 3-connected planar graphs, and IFP+C captures polynomial time on arbitrary planar graphs. (2) Planar graphs can be characterized up to isomorphism in a logic with finitely many variables and counting. This answers a question of Immerman (1987). (3) The class of planar graphs is definable in IFP. This answers a question of Dawar and Gradel.

Far each /spl kappa//spl ges/3, we show that there is no recursive bound for the size of the smallest finite model of an L/sup /spl kappa//-theory in terms of its /spl kappa/-size. Here L/sup /spl kappa// denotes the /spl kappa/-variable fragment of first-order logic. An L/sup /spl kappa//-theory is a maximal consistent set of L/sup /spl kappa//-sentences, and the /spl kappa/-size of an L/sup /spl kappa//-theory is the number of L/sup /spl kappa//-types realized in its models. Our result answers a question of Dawar (1993). As a corollary, we obtain that for /spl kappa//spl ges/3 the so-called L/sup /spl kappa//-invariants, which characterize structures up to equivalence in L/sup /spl kappa//, cannot be recursively inverted.

The main objects of our interest are the existential fragment 9LFP of least xed{point logic, stratied xed point logic SFP, which is the smallest regular logic containing 9LFP, and transitive closure logic TC. The main result of the rst part of this paper is a normal form for 9LFP, which transfers to SFP to a certain extent. We study some of the consequences of this normal form and show that TC can be seen as a natural fragment of SFP. The second part of the paper is concerned with separating the logics under consideration. Furthermore, it shows that the existential preservation theorem fails for TC and SFP (both on nite and arbitrary structures) . The method used to show this also yields a negative answer to a question posed by Rosen and Weinstein [RW95] concerning rst{order sentences preserved under extensions of nite structures. 1 Introduction Inductive denitions by positive existential formulae have rst been studied in generalized recursion theory (see e.g. [Acz77]). Chand...

How difficult is it to decide whether two finite structures can be distinguished in a given logic? For first order logic, this question is equivalent to the graph isomorphism problem with its well-known complexity theoretic difficulties. Somewhat surprisingly, the situation is much clearer when considering the fragments L/sup k/ of first-order logic whose formulae contain at most k (free or bound) variables (for some k/spl ges/1). We show that for each k/spl ges/2, equivalence in the k-variable logic L/sup k/ is complete for polynomial time under quantifier-free reductions (a weak form of NC/sub 0/ reductions). Moreover, we show that the same completeness result holds for the powerful extension C/sup k/ of L/sup k/ with counting quantifiers (for every k/spl ges/2).

The notion of logical reducibilities is derived from the idea of interpretations between theories. It was used by Lovász and Gács [LG77] and Immerman [Imm87] to give complete problems for certain complexity classes and hence establish new connections between logical definability and computational complexity. However, the notion is also interesting in a purely logical context. For example, it is helpful to establish nonexpressibility results. We say that a class of τ-structures is a >complete problem for a logic under L-reductions if it is definable in [τ] and if every class definable in can be ”translated” into by L-formulae (cf. §4). We prove the following theorem: 1.1. Theorem. There are complete problems for partial fixed-point logic and for inductive fixed-point logic under quantifier-free reductions. The main step of the proof is to establish a new normal form for fixed-point formulae (which might be of some interest itself). To obtain this normal form we use theorems of Abiteboul and Vianu [AV91a] that show the equivalence between the fixed-point logics we consider and certain extensions of the database query language Datalog. In [Dah87] Dahlhaus gave a complete problem for least fixed-point logic. Since least fixed-point logic equals inductive fixed-point logic by a well-known result of Gurevich and Shelah [GS86], this already proves one part of our theorem. However, our class gives a natural description of the fixed-point process of an inductive fixed-point formula and hence sheds some light on completely different aspects of the logic than Dahlhaus's construction, which is strongly based on the features of least fixed-point formulae.