We consider the task of out-of-distribution (OOD) generalization, where the distribution shift is due to an unobserved confounder ($Z$) affecting both the covariates ($X$) and the labels ($Y$). This confounding introduces heterogeneity in the predictor, i.e., $P(Y \mid X) = E_{P(Z \mid X)}[P(Y \mid X,Z)]$, making traditional covariate and label shift assumptions unsuitable. OOD generalization differs from traditional domain adaptation in that it does not assume access to the covariate distribution ($X^\text{te}$) of the test samples during training. These conditions create a challenging scenario for OOD robustness: (a) $Z^\text{tr}$ is an unobserved confounder during training, (b) $P^\text{te}(Z) \neq P^\text{tr}(Z)$, (c) $X^\text{te}$ is unavailable during training, and (d) the predictive distribution depends on $P^\text{te}(Z)$. While prior work has developed complex predictors requiring multiple additional variables for identifiability of the latent distribution, we explore a set of identifiability assumptions that yield a surprisingly simple predictor using only a single additional variable. Our approach demonstrates superior empirical performance on several benchmark tasks.

Pre-trained Vision Transformers now serve as powerful tools for computer vision. Yet, efficiently adapting them for multiple tasks remains a challenge that arises from the need to modify the rich hidden representations encoded by the learned weight matrices, without inducing interference between tasks. Current parameter-efficient methods like LoRA, which apply low-rank updates, force tasks to compete within constrained subspaces, ultimately degrading performance. We introduce DiTASK a novel Diffeomorphic Multi-Task Fine-Tuning approach that maintains pre-trained representations by preserving weight matrix singular vectors, while enabling task-specific adaptations through neural diffeomorphic transformations of the singular values. By following this approach, DiTASK enables both shared and task-specific feature modulations with minimal added parameters. Our theoretical analysis shows that DiTASK achieves full-rank updates during optimization, preserving the geometric structure of pretrained features, and establishing a new paradigm for efficient multi-task learning (MTL). Our experiments on PASCAL MTL and NYUD show that DiTASK achieves state-of-the-art performance across four dense prediction tasks, using 75% fewer parameters than existing methods. Our code is available here.

This work introduces Castle, the first framework for schema-only cascade update generation using large language models (LLMs).Despite recent advances in LLMs for Text2SQL code generation, existing approaches focus primarily on SELECT queries, neglecting the challenges of SQL update operations and their ripple effects.Traditional CASCADE UPDATE constraints are static and unsuitable for modern, denormalized databases, which demand dynamic, context-aware updates.Castle enables natural language instructions to trigger multicolumn, causally consistent SQL UPDATE statements, without revealing table content to the model.By framing UPDATE SQL generation as a divide-and-conquer task with LLMs' reasoning capacity, Castle can determine not only which columns must be directly updated, but also how those updates propagate through the schema, causing cascading updates -all via nested queries and substructures that ensure data confidentiality.We evaluate it on realworld causal update scenarios, demonstrating its ability to produce accurate SQL updates, and thereby highlighting the reasoning ability of LLMs in automated DBMS.

Large general purpose pre-trained models have revolutionized computer vision and natural language understanding. However, the development of general purpose pre-trained Graph Neural Networks (GNNs) lags behind other domains due to the lack of suitable generalist node representations. Existing GNN architectures are often tailored to specific task orders, such as node-level, link-level, or higher-order tasks, because different tasks require distinct permutation symmetries, which are difficult to reconcile within a single model. In this paper, we propose _holographic node representations_, a new blueprint for node representations capable of solving tasks of any order. Holographic node representations have two key components: (1) a task-agnostic expansion map, which produces highly expressive, high-dimensional embeddings, free from node-permutation symmetries, to be fed into (2) a reduction map that carefully reintroduces the relevant permutation symmetries to produce low-dimensional, task-specific embeddings. We show that well-constructed expansion maps enable simple and efficient reduction maps, which can be adapted for any task order. Empirical results show that holographic node representations can be effectively pre-trained and reused across tasks of varying orders, yielding up to 100% relative performance improvement, including in cases where prior methods fail entirely.

Traditional Graph Neural Networks (GNNs) cannot generalize to new graphs with node attributes different from the training ones, making zero-shot generalization across different node attribute domains an open challenge in graph machine learning. In this paper, we propose STAGE, which encodes *statistical dependencies* between attributes rather than individual attribute values, which may differ in test graphs. By assuming these dependencies remain invariant under changes in node attributes, STAGE achieves provable generalization guarantees for a family of domain shifts. Empirically, STAGE demonstrates strong zero-shot performance on medium-sized datasets: when trained on multiple graph datasets with different attribute spaces (varying in types and number) and evaluated on graphs with entirely new attributes, STAGE achieves a relative improvement in Hits@1 between 40% to 103% in link prediction and a 10% improvement in node classification compared to state-of-the-art baselines.

Causal discovery from observational data is a fundamental task in artificial intelligence, with far-reaching implications for decision-making, predictions, and interventions. Despite significant advances, existing methods can be broadly categorized as constraint-based or score-based approaches. Constraint-based methods offer rigorous causal discovery but are often hindered by small sample sizes, while score-based methods provide flexible optimization but typically forgo explicit conditional independence testing. This work explores a third avenue: developing differentiable $d$-separation scores, obtained through a percolation theory using soft logic. This enables the implementation of a new type of causal discovery method: gradient-based optimization of conditional independence constraints. Empirical evaluations demonstrate the robust performance of our approach in low-sample regimes, surpassing traditional constraint-based and score-based baselines on a real-world dataset. Code implementing the proposed method is publicly available at [https://github.com/PurdueMINDS/DAGPA](https://github.com/PurdueMINDS/DAGPA).

While recent work shows ML-based approaches are a promising alternative to conventional optimization methods for Traffic Engineering (TE), existing research is limited to a single traffic class. In this paper, we present Hattrick, the first ML-based approach for handling multiple traffic classes, a key requirement of cloud and ISP WANs. As part of Hattrick we have developed (i) a novel neural architecture aligned with the sequence of optimization problems in multiclass TE; and (ii) a variant of classical multitask learning methods to deal with the unique challenge of optimizing multiple metrics that have a precedence relationship. Evaluations on a large private WAN and other public datasets show Hattrick outperforms state-of-the-art optimization-based multiclass TE methods by better coping with prediction error - e.g., for GEANT, Hattrick outperforms SWAN by 5.48% to 19.3% across classes when considering the traffic that can be supported 99% of the time.

In constraint-based causal discovery, the existing algorithms systematically use a series of conditional independence (CI) relations observed in the data to recover an equivalence class of causal graphs in the large sample limit. One limitation of these algorithms is that CI tests lose statistical power as conditioning set size increases with finite samples. Recent research proposes to limit the conditioning set size for robust causal discovery. However, the existing algorithms require exhaustive testing of all CI relations with conditioning set sizes up to a certain integer $k$. This becomes problematic in practice when variables with large support are present, as it makes CI tests less reliable due to near-deterministic relationships, thereby violating the faithfulness assumption. To address this issue, we propose a causal discovery algorithm that only uses CI tests where the conditioning sets are restricted to a given set of conditioning sets including the empty set $\mathcal{C}$. We call such set of CI relations ${\mathcal{I}}_{\mathcal{C}}$ conditionally closed. We define the notion of $\mathcal{C}$-Markov equivalence: two causal graphs are $\mathcal{C}$-Markov equivalent if they entail the same set of CI constraints from ${\mathcal{I}}_\mathcal{C}$. We propose a graphical representation of $\mathcal{C}$-Markov equivalence and characterize such equivalence between two causal graphs. Our proposed algorithm called the $\mathcal{C}$-PC algorithm is sound for learning the $\mathcal{C}$-Markov equivalence class. We demonstrate the utility of the proposed algorithm via synthetic and real-world experiments in scenarios where variables with large support or high correlation are present in the data.

Subgraph GNNs are provably expressive neural architectures that learn graph representations from sets of subgraphs. Unfortunately, their applicability is hampered by the computational complexity associated with performing message passing on many subgraphs. In this paper, we consider the problem of learning to select a small subset of the large set of possible subgraphs in a data-driven fashion. We first motivate the problem by proving that there are families of WL-indistinguishable graphs for which there exist efficient subgraph selection policies: small subsets of subgraphs that can already identify all the graphs within the family. We then propose a new approach, called _Policy-Learn_, that learns how to select subgraphs in an iterative manner. We prove that, unlike popular random policies and prior work addressing the same problem, our architecture is able to learn the efficient policies mentioned above. Our experimental results demonstrate that _Policy-Learn_ outperforms existing baselines across a wide range of datasets.

A fundamental challenge in physics-informed machine learning (PIML) is the design of robust PIML methods for out-of-distribution (OOD) forecasting tasks. These OOD tasks require learning-to-learn from observations of the same (ODE) dynamical system with different unknown ODE parameters, and demand accurate forecasts even under out-of-support initial conditions and out-of-support ODE parameters. In this work we propose to improve the OOD robustness of PIML via a meta-learning procedure for causal structure discovery. Using three different OOD tasks, we empirically observe that the proposed approach significantly outperforms existing state-of-the-art PIML and deep learning methods (with $2\times$ to $28\times$ lower OOD errors).

Complex logical query answering (CLQA) in knowledge graphs (KGs) goes beyond simple KG completion and aims at answering compositional queries comprised of multiple projections and logical operations. Existing CLQA methods that learn parameters bound to certain entity or relation vocabularies can only be applied to the graph they are trained on which requires substantial training time before being deployed on a new graph. Here we present UltraQuery, the first foundation model for inductive reasoning that can zero-shot answer logical queries on any KG. The core idea of UltraQuery is to derive both projections and logical operations as vocabulary-independent functions which generalize to new entities and relations in any KG. With the projection operation initialized from a pre-trained inductive KG completion model, UltraQuery can solve CLQA on any KG after finetuning on a single dataset. Experimenting on 23 datasets, UltraQuery in the zero-shot inference mode shows competitive or better query answering performance than best available baselines and sets a new state of the art on 15 of them.

In this paper, we propose a novel activation function tailored specifically for graph data in Graph Neural Networks (GNNs). Motivated by the need for graph-adaptive and flexible activation functions, we introduce DiGRAF, leveraging Continuous Piecewise-Affine Based (CPAB) transformations, which we augment with an additional GNN to learn a graph-adaptive diffeomorphic activation function in an end-to-end manner. In addition to its graph-adaptivity and flexibility, DiGRAF also possesses properties that are widely recognized as desirable for activation functions, such as differentiability, boundness within the domain, and computational efficiency. We conduct an extensive set of experiments across diverse datasets and tasks, demonstrating a consistent and superior performance of DiGRAF compared to traditional and graph-specific activation functions, highlighting its effectiveness as an activation function for GNNs. Our code is available at https://github.com/ipsitmantri/DiGRAF.

Graph data are inherently complex and heterogeneous, leading to a high natural diversity of distributional shifts. However, it remains unclear how to build machine learning architectures that generalize to the complex distributional shifts naturally occurring in the real world. Here, we develop GraphMETRO, a Graph Neural Network architecture that models natural diversity and captures complex distributional shifts. GraphMETRO employs a Mixture-of-Experts (MoE) architecture with a gating model and multiple expert models, where each expert model targets a specific distributional shift to produce a referential representation w.r.t. a reference model, and the gating model identifies shift components. Additionally, we design a novel objective that aligns the representations from different expert models to ensure reliable optimization. GraphMETRO achieves state-of-the-art results on four datasets from the GOOD benchmark, which is comprised of complex and natural real-world distribution shifts, improving by 67% and 4.2% on the WebKB and Twitch datasets. Code and data are available at https://github.com/Wuyxin/GraphMETRO.

Recently, researchers have proposed ML-driven traffic engineering (TE) schemes where a neural network model is used to produce TE decisions in lieu of conventional optimization solvers. Unfortunately existing ML-based TE schemes are not explicitly designed to be robust to topology changes that may occur due to WAN evolution, failures or planned maintenance. In this paper, we present HARP, a neural model for TE explicitly capable of handling variations in topology including those not observed in training. HARP is designed with two principles in mind: (i) ensure invariances to natural input transformations (e.g., permutations of node ids, tunnel reordering); and (ii) align neural architecture to the optimization model. Evaluations on a multi-week dataset of a large private WAN show HARP achieves an MLU at most 11% higher than optimal over 98% of the time despite encountering significantly different topologies in testing relative to training data. Further, comparisons with state-of-the-art ML-based TE schemes indicate the importance of the mechanisms introduced by HARP to handle topology variability. Finally, when predicted traffic matrices are provided, HARP outperforms classic optimization solvers achieving a median reduction in MLU of 5 to 10% on the true traffic matrix.

There has been a growing excitement that implicit graph generative models could be used to design or discover new molecules for medicine or material design. Because these molecules have not been discovered, they naturally lie in unexplored or scarcely supported regions of the distribution of known molecules. However, prior evaluation methods for implicit graph generative models have focused on validating statistics computed from the thick support (e.g., mean and variance of a graph property). Therefore, there is a mismatch between the goal of generating novel graphs and the evaluation methods. To address this evaluation gap, we design a novel evaluation method called Vertical Validation (VV) that systematically creates thin support regions during the train-test splitting procedure and then reweights generated samples so that they can be compared to the held-out test data. This procedure can be seen as a generalization of the standard train-test procedure except that the splits are dependent on sample features. We demonstrate that our method can be used to perform model selection if performance on thin support regions is the desired goal. As a side benefit, we also show that our approach can better detect overfitting as exemplified by memorization.

In this paper, we consider the task of answering what-if questions in the context of adaptive bit rate (ABR) video streaming without access to randomized control trials (RCTs) (e.g., no A/B testing) - i.e., given recorded data of an existing deployed system, what would be the performance impact if we changed its design. Our work makes three contributions. First, we show the problem is challenging since data may only be available for a single ABR algorithm without RCTs, and since it is necessary to deal with the cascading effects that past ABR decisions have on future decisions. Next we present Veritas, the first framework that tackles causal reasoning for video streaming without requiring data collected through RCTs. Integral to Veritas is an easy-to-interpret domain-specific ML model that relates the latent stochastic process (intrinsic bandwidth that the video session can achieve) to actual observations (download times), while exploiting counterfactual queries via abduction using the observed TCP states (e.g., congestion window) for blocking the cascading dependencies. Third, we evaluate Veritas's ability to accurately answer a wide range of what-if questions using emulation experiments, and data of real video sessions from Puffer. The results show that (i) Veritas accurately tackles a wider range of what-if questions (e.g., change of buffer size or video quality) that existing approaches cannot; (ii) Veritas without RCT training data achieves performance comparable or better than a recent parallel approach that requires RCT data; and (iii) in many scenarios Veritas achieves accuracy close to an ideal oracle.

This work formalizes the associational task of predicting node attribute evolution in temporal graphs from the perspective of learning equivariant representations. We show that node representations in temporal graphs can be cast into two distinct frameworks: (a) The most popular approach, which we denote as time-and-graph, where equivariant graph (e.g., GNN) and sequence (e.g., RNN) representations are intertwined to represent the temporal evolution of node attributes in the graph; and (b) an approach that we denote as time-then-graph, where the sequences describing the node and edge dynamics are represented first, then fed as node and edge attributes into a static equivariant graph representation that comes after. Interestingly, we show that time-then-graph representations have an expressivity advantage over time-and-graph representations when both use component GNNs that are not most-expressive (e.g., 1-Weisfeiler-Lehman GNNs). Moreover, while our goal is not necessarily to obtain state-of-the-art results, our experiments show that time-then-graph methods are capable of achieving better performance and efficiency than state-of-the-art time-and-graph methods in some real-world tasks, thereby showcasing that the time-then-graph framework is a worthy addition to the graph ML toolbox.

This work provides the first theoretical study on the ability of graph Message Passing Neural Networks (gMPNNs) ---such as Graph Neural Networks (GNNs)--- to perform inductive out-of-distribution (OOD) link prediction tasks, where deployment (test) graph sizes are larger than training graphs. We first prove non-asymptotic bounds showing that link predictors based on permutation-equivariant (structural) node embeddings obtained by gMPNNs can converge to a random guess as test graphs get larger. We then propose a theoretically-sound gMPNN that outputs structural pairwise (2-node) embeddings and prove non-asymptotic bounds showing that, as test graphs grow, these embeddings converge to embeddings of a continuous function that retains its ability to predict links OOD. Empirical results on random graphs show agreement with our theoretical results.

We study the membership inference (MI) attack against classifiers, where the attacker's goal is to determine whether a data instance was used for training the classifier. Through systematic cataloging of existing MI attacks and extensive experimental evaluations of them, we find that a model's vulnerability to MI attacks is tightly related to the generalization gap---the difference between training accuracy and test accuracy. We then propose a defense against MI attacks that aims to close the gap by intentionally reduces the training accuracy. More specifically, the training process attempts to match the training and validation accuracies, by means of a new set regularizer using the Maximum Mean Discrepancy between the softmax output empirical distributions of the training and validation sets. Our experimental results show that combining this approach with another simple defense (mix-up training) significantly improves state-of-the-art defense against MI attacks, with minimal impact on testing accuracy.

In general, graph representation learning methods assume that the train and test data come from the same distribution. In this work we consider an underexplored area of an otherwise rapidly developing field of graph representation learning: The task of out-of-distribution (OOD) graph classification, where train and test data have different distributions, with test data unavailable during training. Our work shows it is possible to use a causal model to learn approximately invariant representations that better extrapolate between train and test data. Finally, we conclude with synthetic and real-world dataset experiments showcasing the benefits of representations that are invariant to train/test distribution shifts.

Collective Inference (CI) is a procedure designed to boost weak relational classifiers, specially for node classification tasks. Graph Neural Networks (GNNs) are strong classifiers that have been used with great success. Unfortunately, most existing practical GNNs are not most-expressive (universal). Thus, it is an open question whether one can improve strong relational node classifiers, such as GNNs, with CI. In this work, we investigate this question and propose {\em collective learning} for GNNs —a general collective classification approach for node representation learning that increases their representation power. We show that previous attempts to incorporate CI into GNNs fail to boost their expressiveness because they do not adapt CI’s Monte Carlo sampling to representation learning. We evaluate our proposed framework with a variety of state-of-the-art GNNs. Our experiments show a consistent, significant boost in node classification accuracy —regardless of the choice of underlying GNN— for inductive node classification in partially-labeled graphs, across five real-world network datasets.

Graph neural networks (GNNs) have limited expressive power, failing to represent many graph classes correctly. While more expressive graph representation learning (GRL) alternatives can distinguish some of these classes, they are significantly harder to implement, may not scale well, and have not been shown to outperform well-tuned GNNs in real-world tasks. Thus, devising simple, scalable, and expressive GRL architectures that also achieve real-world improvements remains an open challenge. In this work, we show the extent to which graph reconstruction---reconstructing a graph from its subgraphs---can mitigate the theoretical and practical problems currently faced by GRL architectures. First, we leverage graph reconstruction to build two new classes of expressive graph representations. Secondly, we show how graph reconstruction boosts the expressive power of any GNN architecture while being a (provably) powerful inductive bias for invariances to vertex removals. Empirically, we show how reconstruction can boost GNN's expressive power---while maintaining its invariance to permutations of the vertices---by solving seven graph property tasks not solvable by the original GNN. Further, we demonstrate how it boosts state-of-the-art GNN's performance across nine real-world benchmark datasets.

Image classifiers often suffer from adversarial examples, which are generated by strategically adding a small amount of noise to input images to trick classifiers into misclassification. Over the years, many defense mechanisms have been proposed, and different researchers have made seemingly contradictory claims on their effectiveness. We present an analysis of possible adversarial models, and propose an evaluation framework for comparing different defense mechanisms. As part of the framework, we introduce a more powerful and realistic adversary strategy. Furthermore, we propose a new defense mechanism called Random Spiking (RS), which generalizes dropout and introduces random noises in the training process in a controlled manner. Evaluations under our proposed framework suggest RS delivers better protection against adversarial examples than many existing schemes.

Cellular service carriers often employ reactive strategies to assist customers who experience non-outage related individual service degradation issues (e.g., service performance degradations that do not impact customers at scale and are likely caused by network provisioning issues for individual devices). Customers need to contact customer care to request assistance before these issues are resolved. This paper presents our experience with PACE (ProActive customer CarE), a novel, proactive system that monitors, troubleshoots and resolves individual service issues, without having to rely on customers to first contact customer care for assistance. PACE seeks to improve customer experience and care operation efficiency by automatically detecting individual (non-outage related) service issues, prioritizing repair actions by predicting customers who are likely to contact care to report their issues, and proactively triggering actions to resolve these issues. We develop three machine learning-based prediction models, and implement a fully automated system that integrates these prediction models and takes resolution actions for individual customers. We conduct a large-scale trace-driven evaluation using real-world data collected from a major cellular carrier in the US, and demonstrate that PACE is able to predict customers who are likely to contact care due to non-outage related individual service issues with high accuracy. We further deploy PACE into this cellular carrier network. Our field trial results show that PACE is effective in proactively resolving non-outage related individual customer service issues, improving customer experience, and reducing the need for customers to report their service issues.

Existing Graph Neural Network (GNN) methods that learn inductive unsupervised graph representations focus on learning node and edge representations by predicting observed edges in the graph. Although such approaches have shown advances in downstream node classification tasks, they are ineffective in jointly representing larger k-node sets, k{>}2. We propose MHM-GNN, an inductive unsupervised graph representation approach that combines joint k-node representations with energy-based models (hypergraph Markov networks) and GNNs. To address the intractability of the loss that arises from this combination, we endow our optimization with a loss upper bound using a finite-sample unbiased Markov Chain Monte Carlo estimator. Our experiments show that the unsupervised joint k-node representations of MHM-GNN produce better unsupervised representations than existing approaches from the literature.

This work generalizes graph neural networks (GNNs) beyond those based on the Weisfeiler-Lehman (WL) algorithm, graph Laplacians, and diffusions. Our approach, denoted Relational Pooling (RP), draws from the theory of finite partial exchangeability to provide a framework with maximal representation power for graphs. RP can work with existing graph representation models and, somewhat counterintuitively, can make them even more powerful than the original WL isomorphism test. Additionally, RP allows architectures like Recurrent Neural Networks and Convolutional Neural Networks to be used in a theoretically sound approach for graph classification. We demonstrate improved performance of RP-based graph representations over state-of-the-art methods on a number of tasks.

In many complex domains, the input data are often not suited for the typical vector representations used in deep learning models. For example, in relational learning and computer vision tasks, the data are often better represented as sets (e.g., the neighborhood of a node, a cloud of points). In these cases, a key challenge is to learn an embedding function that is invariant to permutations of the input. While there has been some recent work on principled methods for learning permutation-invariant representations of sets, these approaches are limited in their applicability to set-of-sets (SoS) tasks, such as subgraph prediction and scene classification. In this work, we develop a deep neural network framework to learn inductive SoS embeddings that are invariant to SoS permutations. Specifically, we propose HATS, a hierarchical sequence model with attention mechanisms for inductive set-of-sets embeddings. We develop stochastic optimization and inference methods for learning HATS, and our experiments demonstrate that HATS achieves superior performance across a wide range of set-of-sets tasks.

The lack of diversity in computing is a well-known issue. This poster is a work-in-progress report on Curated Pathways to Innovation (CPI), a web-based tool which gathers existing online resources for computer science (CS) engagement and learning to allow students to learn more about CS careers and content, with a particular focus on improving participation of K-12 girls and under-represented minorities in CS. This project is a collaboration of people from academia in CS and social science, K-12 education, non-profit, and industry. We are about halfway through a 3-year pilot deployment of CPI with all students in a low-income, primarily Latino/a middle school with nearly 500 students, and smaller deployments have been undertaken and are planned for 2018-19. In addition to online content, we have created in-person experiences, including reverse science fairs, summer camps, and a hackathon, which are tracked in the CPI tool. To measure impact, we conduct regular surveys with the students measuring their interest in CS, self-efficacy, and other metrics. Our evaluation of the system based on survey data has helped inform the development of the system and curriculum, but remains preliminary. This poster also discusses the tool itself. It uses gamification in the form of badges to measure student progress. From the beginning, the vision was to use machine learning to customize recommendations based on students' demographics, background, and past performance. This integration is coming to fruition at the same time we are including more interesting visuals in the UI, such as an avatar and animations.

Estimating distributions of node characteristics (labels) such as number of connections or citizenship of users in a social network via edge and node sampling is a vital part of the study of complex networks. Due to its low cost, sampling via a random walk (RW) has been proposed as an attractive solution to this task. Most RW methods assume either that the network is undirected or that walkers can traverse edges regardless of their direction. Some RW methods have been designed for directed networks where edges coming into a node are not directly observable. In this work, we propose Directed Unbiased Frontier Sampling (DUFS), a sampling method based on a large number of coordinated walkers, each starting from a node chosen uniformly at random. It applies to directed networks with invisible incoming edges because it constructs, in real time, an undirected graph consistent with the walkers trajectories, and its use of random jumps to prevent walkers from being trapped. DUFS generalizes previous RW methods and is suited for undirected networks and to directed networks regardless of in-edge visibility. We also propose an improved estimator of node label distribution that combines information from initial walker locations with subsequent RW observations. We evaluate DUFS, compare it to other RW methods, investigate the impact of its parameters on estimation accuracy and provide practical guidelines for choosing them. In estimating out-degree distributions, DUFS yields significantly better estimates of the head of the distribution than other methods, while matching or exceeding estimation accuracy of the tail. Last, we show that DUFS outperforms uniform sampling when estimating distributions of node labels of the top 10% largest degree nodes, even when sampling a node uniformly has the same cost as RW steps.

In this work we generalize traditional node/link prediction tasks in dynamic heterogeneous networks, to consider joint prediction over larger k-node induced subgraphs. Our key insight is to incorporate the unavoidable dependencies in the training observations of induced subgraphs into both the input features and the model architecture itself via high-order dependencies. The strength of the representation is its invariance to isomorphisms and varying local neighborhood sizes, while still being able to take node/edge labels into account, and facilitating inductive reasoning (i.e., generalization to unseen portions of the network). Empirical results show that our proposed method significantly outperforms other state-of-the-art methods designed for static and/or single node/link prediction tasks. In addition, we show that our method is scalable and learns interpretable parameters.

We propose a Las Vegas transformation of Markov Chain Monte Carlo (MCMC) estimators of Restricted Boltzmann Machines (RBMs). We denote our approach Markov Chain Las Vegas (MCLV). MCLV gives statistical guarantees in exchange for random running times. MCLV uses a stopping set built from the training data and has maximum number of Markov chain steps K (referred as MCLV-K). We present a MCLV-K gradient estimator (LVS-K) for RBMs and explore the correspondence and differences between LVS-K and Contrastive Divergence (CD-K), with LVS-K significantly outperforming CD-K training RBMs over the MNIST dataset, indicating MCLV to be a promising direction in learning generative models.

In this work we propose R-GPM, a parallel computing framework for graph pattern mining (GPM) through a user-defined subgraph relation. More specifically, we enable the computation of statistics of patterns through their subgraph classes, generalizing traditional GPM methods. R-GPM provides efficient estimators for these statistics by employing a MCMC sampling algorithm combined with several optimizations. We provide both theoretical guarantees and empirical evaluations of our estimators in application scenarios such as stochastic optimization of deep high-order graph neural network models and pattern (motif) counting. We also propose and evaluate optimizations that enable improvements of our estimators accuracy, while reducing their computational costs in up to 3-orders-of-magnitude. Finally, we show that R-GPM is scalable, providing near-linear speedups.

Most content providers are interested in providing good video delivery QoE for all users, not just on average. State-of-the-art ABR algorithms like BOLA and MPC rely on parameters that are sensitive to network conditions, so may perform poorly for some users and/or videos. In this paper, we propose a technique called Oboe to auto-tune these parameters to different network conditions. Oboe pre-computes, for a given ABR algorithm, the best possible parameters for different network conditions, then dynamically adapts the parameters at run-time for the current network conditions. Using testbed experiments, we show that Oboe significantly improves BOLA, MPC, and a commercially deployed ABR. Oboe also betters a recently proposed reinforcement learning based ABR, Pensieve, by 24% on average on a composite QoE metric, in part because it is able to better specialize ABR behavior across different network states.

Applications in various domains rely on processing graph streams, e.g., communication logs of a cloud-troubleshooting system, road-network traffic updates, and interactions on a social network. A labeled-graph stream refers to a sequence of streamed edges of distinct types that form a labeled graph. Due to the large volume and high velocity of these streams, it is often more practical to incrementally build a lossy-compressed version of the graph, and use this lossy version to approximately evaluate graph queries. Challenges arise when the queries are unknown in advance but are associated with filtering predicates based on edge labels. Surprisingly common, and especially challenging, are labeled-graph streams that have highly skewed and unpredictable label-distributions. This paper introduces Self-Balanced Graph Sketch (SBG-Sketch, for short), a graph sketch for summarizing and querying labeled-graph streams, coping with highly imbalanced labels. SBG-Sketch maintains synopsis for both the edge attributes as well as the topology of the streamed graph. SBG-Sketch allows efficient processing of traversal queries, e.g., reachability queries. Experimental results over a variety of real labeled-graph streams show SBG-Sketch to reduce the estimation errors of state-of-the-art methods by up to 99%.

Are Online Social Network (OSN) A users more likely to form friendships with those with similar attributes? Do users at an OSN B score content more favorably than OSN C users? Such questions frequently arise in the context of Social Network Analysis (SNA) but often crawling an OSN network via its Application Programming Interface (API) is the only way to gather data from a third party. To date, these partial API crawls are the majority of public datasets and the synonym of lack of statistical guarantees in incomplete-data comparisons, severely limiting SNA research progress. Using regenerative properties of the random walks, we propose estimation techniques based on short crawls that have proven statistical guarantees. Moreover, our short crawls can be implemented in massively distributed algorithms. We also provide an adaptive crawler that makes our method parameter-free, significantly improving our statistical guarantees. We then derive the Bayesian approximation of the posterior of the estimates, and in addition, obtain an estimator for the expected value of node and edge statistics in an equivalent configuration model or Chung-Lu random graph model of the given network (where nodes are connected randomly) and use it as a basis for testing null hypotheses. The theoretical results are supported with simulations on a variety of real-world networks.

Cumulative advantage (CA) refers to the notion that accumulated resources foster the accumulation of further resources in competitions, a phenomenon that has been empirically observed in various contexts. The oldest and arguably simplest mathematical model that embodies this general principle is the Pólya urn process, which finds applications in a myriad of problems. The original model captures the dynamics of competitions between two equally fit agents under linear CA effects, which can be readily generalized to incorporate different fitnesses and nonlinear CA effects. We study two statistics of competitions under the generalized model, namely duration (i.e., time of the last tie) and intensity (i.e., number of ties). We give rigorous mathematical characterizations of the tail distributions of both duration and intensity under the various regimes for fitness and nonlinearity, which reveal very interesting behaviors. For example, fitness superiority induces much shorter competitions in the sublinear regime while much longer competitions in the superlinear regime. Our findings can shed light on the application of Pólya urn processes in more general contexts where fitness and nonlinearity may be present.

Which song will Smith listen to next? Which restaurant will Alice go to tomorrow? Which product will John click next? These applications have in common the prediction of user trajectories that are in a constant state of flux over a hidden network (e.g. website links, geographic location). Moreover, what users are doing now may be unrelated to what they will be doing in an hour from now. Mindful of these challenges we propose TribeFlow, a method designed to cope with the complex challenges of learning personalized predictive models of non-stationary, transient, and time-heterogeneous user trajectories. TribeFlow is a general method that can perform next product recommendation, next song recommendation, next location prediction, and general arbitrary-length user trajectory prediction without domain-specific knowledge. TribeFlow is more accurate and up to 413x faster than top competitors.

How does a new startup drive the popularity of competing websites into oblivion like Facebook famously did to MySpace? This question is of great interest to academics, technologists, and financial investors alike. In this work we exploit the singular way in which Facebook wiped out the popularity of MySpace, Hi5, Friendster, and Multiply to guide the design of a new popularity competition model. Our model provides new insights into what Nobel Laure- ate Herbert A. Simon called the "marketplace of attention," which we recast as the attention-activity marketplace. Our model design is further substantiated by user-level activity of 250,000 MySpace users obtained between 2004 and 2009. The resulting model not only accurately fits the observed Daily Active Users (DAU) of Facebook and its competitors but also predicts their fate four years into the future.

Complex network phenomena -- such as information cascades in online social networks -- are hard to fully observe, model, and forecast. In forecasting, a recent trend has been to forgo the use of parsimonious models in favor of models with increasingly large degrees of freedom that are trained to learn the behavior of a process from historical data. Extrapolating this trend into the future, eventually we would renounce models all together. But is it possible to forecast the evolution of a complex stochastic process directly from the data without a model? In this work we show that model-free forecasting is possible. We present SED, an algorithm that forecasts process statistics based on relationships of statistical equivalence using two general axioms and historical data. To the best of our knowledge, SED is the first method that can perform axiomatic, model-free forecasts of complex stochastic processes. Our simulations using simple and complex evolving processes and tests performed on a large real-world dataset show promising results.

Driven by outstanding success stories of Internet startups such as Facebook and The Huffington Post, recent studies have thoroughly described their growth. These highly visible online success stories, however, overshadow an untold number of similar ventures that fail. The study of website popularity is ultimately incomplete without general mechanisms that can describe both successes and failures. In this work we present six years of the daily number of users (DAU) of twenty-two membership-based websites - encompassing online social networks, grassroots movements, online forums, and membership-only Internet stores - well balanced between successes and failures. We then propose a combination of reaction-diffusion-decay processes whose resulting equations seem not only to describe well the observed DAU time series but also provide means to roughly predict their evolution. This model allows an approximate automatic DAU-based classification of websites into self-sustainable v.s. unsustainable and whether the startup growth is mostly driven by marketing & media campaigns or word-of-mouth adoptions.

Exploring statistics of locally connected subgraph patterns (also known as network motifs) has helped researchers better understand the structure and function of biological and Online Social Networks (OSNs). Nowadays, the massive size of some critical networks—often stored in already overloaded relational databases—effectively limits the rate at which nodes and edges can be explored, making it a challenge to accurately discover subgraph statistics. In this work, we propose sampling methods to accurately estimate subgraph statistics from as few queried nodes as possible. We present sampling algorithms that efficiently and accurately estimate subgraph properties of massive networks. Our algorithms require no precomputation or complete network topology information. At the same time, we provide theoretical guarantees of convergence. We perform experiments using widely known datasets and show that, for the same accuracy, our algorithms require an order of magnitude less queries (samples) than the current state-of-the-art algorithms.

Online dating sites have become popular platforms for people to look for potential romantic partners. It is important to understand users' dating preferences in order to make better recommendations on potential dates. The message sending and replying actions of a user are strong indicators for what he/she is looking for in a potential date and reflect the user's actual dating preferences. We study how users' online dating behaviors correlate with various user attributes using a real-world dateset from a major online dating site in China. Our study provides a firsthand account of the user online dating behaviors in China, a country with a large population and unique culture. The results can provide valuable guidelines to the design of recommendation engine for potential dates.

In this work we study the set size distribution estimation problem, where elements are randomly sampled from a collection of non-overlapping sets and we seek to recover the original set size distribution from the samples. This problem has applications to capacity planning and network theory. Examples of real-world applications include characterizing in-degree distributions in large graphs and uncovering TCP/IP flow size distributions on the Internet. We demonstrate that it is difficult to estimate the original set size distribution. The recoverability of original set size distributions presents a sharp threshold with respect to the fraction of elements that remain in the sets. If this fraction lies below the threshold, typically half of the elements in power-law and heavier-than-exponential-tailed distributions, then the original set size distribution is unrecoverable. We also discuss practical implications of our findings.

Despite recent efforts to characterize complex networks such as citation graphs or online social networks (OSNs), little attention has been given to developing tools that can be used to characterize directed graphs in the wild, where no pre-processed data is available. The presence of hidden incoming edges but observable outgoing edges poses a challenge to characterize large directed graphs through crawling, as existing sampling methods cannot cope with hidden incoming links. The driving principle behind our random walk (RW) sampling method is to construct, in real-time, an undirected graph from the directed graph such that the random walk on the directed graph is consistent with one on the undirected graph. We then use the RW on the undirected graph to estimate the outdegree distribution. Our algorithm accurately estimates outdegree distributions of a variety of real world graphs. We also study the hardness of indegree distribution estimation when indegrees are latent (i.e., incoming links are only observed as outgoing edges). We observe that, in the same scenarios, indegree distribution estimates are highly innacurate unless the directed graph is highly symmetrical.

In this paper we study the behavior of a continuous time random walk (CTRW) on a stationary and ergodic time varying dynamic graph. We establish conditions under which the CTRW is a stationary and ergodic process. In general, the stationary distribution of the walker depends on the walker rate and is difficult to characterize. However, we characterize the stationary distribution in the following cases: i) the walker rate is significantly larger or smaller than the rate in which the graph changes (time-scale separation), ii) the walker rate is proportional to the degree of the node that it resides on (coupled dynamics), and iii) the degrees of node belonging to the same connected component are identical (structural constraints). We provide examples that illustrate our theoretical findings.

No abstract available.

Estimating characteristics of large graphs via sampling is a vital part of the study of complex networks. Current sampling methods such as (independent) random vertex and random walks are useful but have drawbacks. Random vertex sampling may require too many resources (time, bandwidth, or money). Random walks, which normally require fewer resources per sample, can suffer from large estimation errors in the presence of disconnected or loosely connected graphs. In this work we propose a new m-dimensional random walk that uses m dependent random walkers. We show that the proposed sampling method, which we call Frontier sampling, exhibits all of the nice sampling properties of a regular random walk. At the same time, our simulations over large real world graphs show that, in the presence of disconnected or loosely connected components, Frontier sampling exhibits lower estimation errors than regular random walks. We also show that Frontier sampling is more suitable than random vertex sampling to sample the tail of the degree distribution of the graph.

The histogram of network flow sizes is an important yet difficult metric to estimate in network monitoring. It is important because it characterizes traffic compositions and is a crucial component of anomaly detection methods. It is difficult to estimate because of its high memory and computational requirements. Existing algorithms compute fine grained estimates for each flow size, i.e. 1, 2,... up to the maximum number observed over a finite time interval. Our approach instead relies on the insight that, while many applications require fine grained estimates of small flow sizes, i.e. {1,2,...,k} with a small k, network operators are often only interested in coarse grained estimates of larger flow sizes. Thus, we propose an estimator that outputs a binned histogram of size distributions. Our estimator computes this histogram in O(k3 + log W) operations, where W is the largest flow size of interest to the network operator, while requiring only a few bits of memory per measured flow. This translates into more than 4 fold memory savings and an exponential speedup in the estimator as compared to previous works, greatly increasing the possibility of performing on-line estimation inside a router.

Packet sampling is widely used in network monitoring. Sampled packet streams are often used to determine flow-level statistics of network traffic. To date there is conflicting evidence on the quality of the resulting estimates. In this paper we take a systematic approach, using the Fisher information metric and the Cramér-Rao bound, to understand the contributions that different types of information within sampled packets have on the quality of flow-level estimates. We provide concrete evidence that, without protocol information and with packet sampling rate p = 0.005, any accurate unbiased estimator needs approximately 1016 sampled flows. The required number of sampled flows drops to roughly 104 with the use of TCP sequence numbers. Furthermore, additional SYN flag information significantly reduces the estimation error of short flows. We present a Maximum Likelihood Estimator (MLE) that relies on all of this information and show that it is efficient, even when applied to a small sample set. We validate our results using Tier-1 Internet backbone traces and evaluate the benefits of sampling from multiple monitors. Our results show that combining estimates from several monitors is 50% less accurate than an estimate based on all samples.