Emerging optical switching technologies enable demand-aware datacenter networks, whose topology can be flexibly optimized toward the traffic they serve. This paper revisits the bounded-degree network design problem underlying such demand-aware networks. Namely, given a distribution over communicating node pairs (represented has a demand graph), we want to design a network with bounded maximum degree (called host graph) that minimizes the expected communication distance. We improve the understanding of this problem domain by filling several gaps in prior work. First, we present the first practical algorithm for solving this problem on arbitrary instances without violating the degree bound. Our algorithm is based on novel insights obtained from studying a new Steiner node version of the problem, and we report on an extensive empirical evaluation, using several real-world traffic traces from datacenters, finding that our approach results in improved demand-aware network designs. Second, we shed light on the complexity and hardness of the bounded-degree network design problem by formally establishing its NP-completeness for any degree. We use our techniques to improve prior upper bounds for sparse instances. Finally, we study an intriguing connection between demand-aware network design and the virtual networking embedding problem, and show that the latter cannot be used to approximate the former: there is no universal host graph which can provide a constant approximation for our problem.

In prior work, Gupta et al. (SPAA 2022) presented a distributed algorithm for multiplying sparse n x n matrices, using n computers. They assumed that the input matrices are uniformly sparse---there are at most d non-zeros in each row and column---and the task is to compute a uniformly sparse part of the product matrix. Initially each computer knows one row of each input matrix, and eventually each computer needs to know one row of the product matrix. In each communication round each computer can send and receive one O(łog n)-bit message. Their algorithm solves this task in O(d^1.907 ) rounds, while the trivial bound is O(d^2).

We study matrix multiplication in the low-bandwidth model: There are n computers, and we need to compute the product of two n × n matrices. Initially computer i knows row i of each input matrix. In one communication round each computer can send and receive one O(logn)-bit message. Eventually computer i has to output row i of the product matrix.

We consider a standard distributed optimisation setting where $N$ machines, each holding a $d$-dimensional function $f_i$, aim to jointly minimise the sum of the functions $\sum_{i = 1}^N f_i (x)$. This problem arises naturally in large-scale distributed optimisation, where a standard solution is to apply variants of (stochastic) gradient descent. We focus on the communication complexity of this problem: our main result provides the first fully unconditional bounds on total number of bits which need to be sent and received by the $N$ machines to solve this problem under point-to-point communication, within a given error-tolerance. Specifically, we show that $\Omega( Nd \log d / N\varepsilon)$ total bits need to be communicated between the machines to find an additive $\epsilon$-approximation to the minimum of $\sum_{i = 1}^N f_i (x)$. The result holds for both deterministic and randomised algorithms, and, importantly, requires no assumptions on the algorithm structure. The lower bound is tight under certain restrictions on parameter values, and is matched within constant factors for quadratic objectives by a new variant of quantised gradient descent, which we describe and analyse. Our results bring over tools from communication complexity to distributed optimisation, which has potential for further applications.

Subgraph detection has recently been one of the most studied problems in the CONGEST model of distributed computing. In this work, we study the distributed complexity of problems closely related to subgraph detection, mainly focusing on induced subgraph detection. The main line of this work presents lower bounds and parameterized algorithms w.r.t structural parameters of the input graph: - On general graphs, we give unconditional lower bounds for induced detection of cycles and patterns of treewidth 2 in CONGEST. Moreover, by adapting reductions from centralized parameterized complexity, we prove lower bounds in CONGEST for detecting patterns with a 4-clique, and for induced path detection conditional on the hardness of triangle detection in the congested clique. - On graphs of bounded degeneracy, we show that induced paths can be detected fast in CONGEST using techniques from parameterized algorithms, while detecting cycles and patterns of treewidth 2 is hard. - On graphs of bounded vertex cover number, we show that induced subgraph detection is easy in CONGEST for any pattern graph. More specifically, we adapt a centralized parameterized algorithm for a more general maximum common induced subgraph detection problem to the distributed setting. In addition to these induced subgraph detection results, we study various related problems in the CONGEST and congested clique models, including for multicolored versions of subgraph-detection-like problems.

We show that any algorithm that solves the sinkless orientation problem in the supported LOCAL model requires Ω(log n) rounds, and this is tight. The supported LOCAL is at least as strong as the usual LOCAL model, and as a corollary this also gives a new, short and elementary proof that shows that the round complexity of the sinkless orientation problem in the deterministic LOCAL model is Ω(log n).

Abstract We design fast deterministic algorithms for distance computation in the Congested Clique model. Our key contributions include: A $$(2+\epsilon )$$ ( 2 + ϵ ) -approximation for all-pairs shortest paths in $$O(\log ^2{n} / \epsilon )$$ O ( log 2 n / ϵ ) rounds on unweighted undirected graphs. With a small additional additive factor, this also applies for weighted graphs. This is the first sub-polynomial constant-factor approximation for APSP in this model. A $$(1+\epsilon )$$ ( 1 + ϵ ) -approximation for multi-source shortest paths from $$O(\sqrt{n})$$ O ( n ) sources in $$O(\log ^2{n} / \epsilon )$$ O ( log 2 n / ϵ ) rounds on weighted undirected graphs. This is the first sub-polynomial algorithm obtaining this approximation for a set of sources of polynomial size. Our main techniques are new distance tools that are obtained via improved algorithms for sparse matrix multiplication, which we leverage to construct efficient hopsets and shortest paths. Furthermore, our techniques extend to additional distance problems for which we improve upon the state-of-the-art, including diameter approximation, and an exact single-source shortest paths algorithm for weighted undirected graphs in $$\tilde{O}(n^{1/6})$$ O ~ ( n 1 / 6 ) rounds.

The ability to leverage large-scale hardware parallelism has been one of the key enablers of the accelerated recent progress in machine learning. Consequently, there has been considerable effort invested into developing efficient parallel variants of classic machine learning algorithms. However, despite the wealth of knowledge on parallelization, some classic machine learning algorithms often prove hard to parallelize efficiently while maintaining convergence. In this paper, we focus on efficient parallel algorithms for the key machine learning task of inference on graphical models, in particular on the fundamental belief propagation algorithm. We address the challenge of efficiently parallelizing this classic paradigm by showing how to leverage scalable relaxed schedulers, which reduce parallelization overheads, in this context. We investigate the overheads of relaxation analytically, and present an extensive empirical study, showing that our approach outperforms previous parallel belief propagation implementations both in terms of scalability and in terms of wall-clock convergence time, on a range of practical applications.

This paper investigates the power of preprocessing in the CONGEST model. Schmid and Suomela (ACM HotSDN 2013) introduced the SUPPORTED CONGEST model to study the application of distributed algorithms in Software-Defined Networks (SDNs). In this paper, we show that a large class of lower bounds in the CONGEST model still hold in the SUPPORTED model, highlighting the robustness of these bounds. This also raises the question how much does preprocessing help in the CONGEST model

The congested clique model of distributed computing has been receiving attention as a model for densely connected distributed systems. While there has been significant progress on the side of upper bounds, we have very little in terms of lower bounds for the congested clique; indeed, it is now known that proving explicit congested clique lower bounds is as difficult as proving circuit lower bounds. In this work, we use various more traditional complexity theory tools to build a clearer picture of the complexity landscape of the congested clique: \beginitemize ıtem Nondeterminism and beyond: We introduce the nondeterministic congested clique model (analogous to NP) and show that there is a natural canonical problem family that captures all problems solvable in constant time with nondeterministic algorithms. We further generalise these notions by introducing the constant-round decision hierarchy (analogous to the polynomial hierarchy). ıtem Non-constructive lower bounds: We lift the prior non-uniform counting arguments to a general technique for proving non-constructive uniform lower bounds for the congested clique. In particular, we prove a time hierarchy theorem for the congested clique, showing that there are decision problems of essentially all complexities, both in the deterministic and nondeterministic settings. ıtem Fine-grained complexity: We map out relationships between various natural problems in the congested clique model, arguing that a reduction-based complexity theory currently gives us a fairly good picture of the complexity landscape of the congested clique. \enditemize

A number of recent papers – e.g. Brandt et al. (STOC 2016), Chang et al. (FOCS 2016), Ghaffari & Su (SODA 2017), Brandt et al. (PODC 2017), and Chang & Pettie (FOCS 2017) – have advanced our understanding of one of the most fundamental questions in theory of distributed computing: what are the possible time complexity classes of LCL problems in the LOCAL model? In essence, we have a graph problem Π in which a solution can be verified by checking all radius-O(1) neighbourhoods, and the question is what is the smallest T such that a solution can be computed so that each node chooses its own output based on its radius-T neighbourhood. Here T is the distributed time complexity of Π. The time complexity classes for deterministic algorithms in bounded-degree graphs that are known to exist by prior work are Θ(1), Θ(log* n), Θ(logn), Θ(n1/k), and Θ(n). It is also known that there are two gaps: one between ω(1) and o(loglog* n), and another between ω(log* n) and o(logn). It has been conjectured that many more gaps exist, and that the overall time hierarchy is relatively simple – indeed, this is known to be the case in restricted graph families such as cycles and grids. We show that the picture is much more diverse than previously expected. We present a general technique for engineering LCL problems with numerous different deterministic time complexities, including Θ(logα n) for any α ≥ 1, 2Θ(logα n) for any α ≤ 1, and Θ(nα) for any α < 1/2 in the high end of the complexity spectrum, and Θ(logα log* n) for any α ≥ 1, 2Θ(logα log* n) for any α ≤ 1, and Θ((log* n)α) for any α ≤ 1 in the low end of the complexity spectrum; here α is a positive rational number.

Developing accurate algorithms for learning structures of probabilistic graphical models is an important problem within modern AI research. Here we focus on score-based structure learning for Bayesian networks as arguably the most central class of graphical models. A successful generic approach to optimal Bayesian network structure learning (BNSL), based on integer programming (IP), is implemented in the Gobnilp system. Despite the recent algorithmic advances, current understanding of foundational aspects underlying the IP based approach to BNSL is still somewhat lacking. In this paper, we provide theoretical contributions towards understanding fundamental aspects of cutting planes and the related separation problem in this context, ranging from NP-hardness results to analysis of polytopes and the related facets in connection to BNSL.

We present simple deterministic algorithms for subgraph finding and enumeration in the broadcast CONGEST model of distributed computation: - For any constant k, detecting k-paths and trees on k nodes can be done in O(1) rounds. - For any constant k, detecting k-cycles and pseudotrees on k nodes can be done in O(n) rounds. - On d-degenerate graphs, cliques and 4-cycles can be enumerated in O(d + log n) rounds, and 5-cycles in O(d2 + log n) rounds. In many cases, these bounds are tight up to logarithmic factors. Moreover, we show that the algorithms for d-degenerate graphs can be improved to O(d/logn) and O(d2/logn), respect- ively, in the supported CONGEST model, which can be seen as an intermediate model between CONGEST and the congested clique.

LCLs or locally checkable labelling problems (e.g. maximal independent set, maximal matching, and vertex colouring) in the LOCAL model of computation are very well-understood in cycles (toroidal 1-dimensional grids): every problem has a complexity of O(1), Θ(log* n), or Θ(n), and the design of optimal algorithms can be fully automated. This work develops the complexity theory of LCL problems for toroidal 2-dimensional grids. The complexity classes are the same as in the 1-dimensional case: O(1), Θ(log* n), and Θ(n). However, given an LCL problem it is undecidable whether its complexity is Θ(log* n) or Θ(n) in 2-dimensional grids.

The congested clique model of distributed computing has been receiving attention as a model for densely connected distributed systems. While there has been significant progress on the side of upper bounds, we have very little in terms of lower bounds for the congested clique; indeed, it is now know that proving explicit congested clique lower bounds is as difficult as proving circuit lower bounds. In this work, we use traditional complexity-theoretic tools to build a clearer picture of the complexity landscape of the congested clique, proving non-constructive lower bounds and studying the relationships between natural problems.

Motivation: While the position weight matrix (PWM) is the most popular model for sequence motifs, there is growing evidence of the usefulness of more advanced models such as first-order Markov representations, and such models are also becoming available in well-known motif databases. There has been lots of research of how to learn these models from training data but the problem of predicting putative sites of the learned motifs by matching the model against new sequences has been given less attention. Moreover, motif site analysis is often concerned about how different variants in the sequence affect the sites. So far, though, the corresponding efficient software tools for motif matching have been lacking. Results: We develop fast motif matching algorithms for the aforementioned tasks. First, we formalize a framework based on high-order position weight matrices for generic representation of motif models with dinucleotide or general q -mer dependencies, and adapt fast PWM matching algorithms to the high-order PWM framework. Second, we show how to incorporate different types of sequence variants , such as SNPs and indels, and their combined effects into efficient PWM matching workflows. Benchmark results show that our algorithms perform well in practice on genome-sized sequence sets and are for multiple motif search much faster than the basic sliding window algorithm. Availability and Implementation: Implementations are available as a part of the MOODS software package under the GNU General Public License v3.0 and the Biopython license ( http://www.cs.helsinki.fi/group/pssmfind ). Contact: [email protected].

The challenging task of learning structures of probabilistic graphical models is an important problem within modern AI research. Recent years have witnessed several major algorithmic advances in structure learning for Bayesian networks - arguably the most central class of graphical models - especially in what is known as the score-based setting. A successful generic approach to optimal Bayesian network structure learning (BNSL), based on integer programming (IP), is implemented in the GOBNILP system. Despite the recent algorithmic advances, current understanding of foundational aspects underlying the IP based approach to BNSL is still somewhat lacking. Understanding fundamental aspects of cutting planes and the related separation problem is important not only from a purely theoretical perspective, but also since it holds out the promise of further improving the efficiency of state-of-the-art approaches to solving BNSL exactly. In this paper, we make several theoretical contributions towards these goals: (i) we study the computational complexity of the separation problem, proving that the problem is NP-hard; (ii) we formalise and analyse the relationship between three key polytopes underlying the IP-based approach to BNSL; (iii) we study the facets of the three polytopes both from the theoretical and practical perspective, providing, via exhaustive computation, a complete enumeration of facets for low-dimensional family-variable polytopes; and, furthermore, (iv) we establish a tight connection of the BNSL problem to the acyclic subgraph problem.

Both learning and inference tasks on Bayesian networks are NP-hard in general. Bounded tree-width Bayesian networks have recently received a lot of attention as a way to circumvent this complexity issue; however, while inference on bounded tree-width networks is tractable, the learning problem remains NP-hard even for tree-width~2. In this paper, we propose bounded vertex cover number Bayesian networks as an alternative to bounded tree-width networks. In particular, we show that both inference and learning can be done in polynomial time for any fixed vertex cover number bound $k$, in contrast to the general and bounded tree-width cases; on the other hand, we also show that learning problem is W[1]-hard in parameter $k$. Furthermore, we give an alternative way to learn bounded vertex cover number Bayesian networks using integer linear programming (ILP), and show this is feasible in practice.

In this work, we use algebraic methods for studying distance computation and subgraph detection tasks in the congested clique model. Specifically, we adapt parallel matrix multiplication implementations to the congested clique, obtaining an O(n1-2/ω) round matrix multiplication algorithm, where ω < 2.3728639 is the exponent of matrix multiplication. In conjunction with known techniques from centralised algorithmics, this gives significant improvements over previous best upper bounds in the congested clique model. The highlight results include: triangle and 4-cycle counting in O(n0.158) rounds, improving upon the O(n1/3) triangle counting algorithm of Dolev et al. [DISC 2012], a (1 + o(1))-approximation of all-pairs shortest paths in O(n0.158) rounds, improving upon the ~O (n1/2)-round (2 + o(1))-approximation algorithm of Nanongkai [STOC 2014], and computing the girth in O(n0.158) rounds, which is the first non-trivial solution in this model. In addition, we present a novel constant-round combinatorial algorithm for detecting 4-cycles.

Inference in Bayesian networks is known to be NP-hard, but if the network has bounded tree-width, then inference becomes tractable. Not surprisingly, learning networks that closely match the given data and have a bounded tree-width has recently attracted some attention. In this paper we aim to lay groundwork for future research on the topic by studying the exact complexity of this problem. We give the first non-trivial exact algorithm for the NP-hard problem of finding an optimal Bayesian network of tree-width at most w, with running time 3^n n^w + O(1), and provide an implementation of this algorithm. Additionally, we propose a variant of Bayesian network learning with “super-structures”, and show that finding a Bayesian network consistent with a given super-structure is fixed-parameter tractable in the tree-width of the super-structure.

UNLABELLED: MOODS (MOtif Occurrence Detection Suite) is a software package for matching position weight matrices against DNA sequences. MOODS implements state-of-the-art online matching algorithms, achieving considerably faster scanning speed than with a simple brute-force search. MOODS is written in C++, with bindings for the popular BioPerl and Biopython toolkits. It can easily be adapted for different purposes and integrated into existing workflows. It can also be used as a C++ library. AVAILABILITY: The package with documentation and examples of usage is available at http://www.cs.helsinki.fi/group/pssmfind. The source code is also available under the terms of a GNU General Public License (GPL).