arXiv subjects

Marta Kwiatkowska

Publications and source records attributed to Marta Kwiatkowska.

At least 73 records · Page 4Linked to original sources

arXiv2022

Correlated Equilibria and Fairness in Concurrent Stochastic Games

Game-theoretic techniques and equilibria analysis facilitate the design and verification of competitive systems. While algorithmic complexity of equilibria computation has been extensively studied, practical implementation and application of game-theoretic methods is more recent. Tools such as PRISM-games support automated verification and synthesis of zero-sum and (epsilon-optimal subgame-perfect) social welfare Nash equilibria properties for concurrent stochastic games. However, these methods become inefficient as the number of agents grows and may also generate equilibria that yield significant variations in the outcomes for individual agents. Instead, we consider correlated equilibria, in which players can coordinate through public signals, and introduce an alternative optimality criterion of social fairness, which can be applied to both Nash and correlated equilibria. We show that correlated equilibria are easier to compute, are more equitable, and can also improve joint outcomes. We implement algorithms for both normal form games and the more complex case of multi-player concurrent stochastic games with temporal logic specifications.

cs.GT↗

arXiv2022

The King is Naked: on the Notion of Robustness for Natural Language Processing

There is growing evidence that the classical notion of adversarial robustness originally introduced for images has been adopted as a de facto standard by a large part of the NLP research community. We show that this notion is problematic in the context of NLP as it considers a narrow spectrum of linguistic phenomena. In this paper, we argue for semantic robustness, which is better aligned with the human concept of linguistic fidelity. We characterize semantic robustness in terms of biases that it is expected to induce in a model. We study semantic robustness of a range of vanilla and robustly trained architectures using a template-based generative test bed. We complement the analysis with empirical evidence that, despite being harder to implement, semantic robustness can improve performance %gives guarantees for on complex linguistic phenomena where models robust in the classical sense fail.

cs.CL↗

arXiv2021

A Language for Modeling And Optimizing Experimental Biological Protocols

Automation is becoming ubiquitous in all laboratory activities, leading towards precisely defined and codified laboratory protocols. However, the integration between laboratory protocols and mathematical models is still lacking. Models describe physical processes, while protocols define the steps carried out during an experiment: neither cover the domain of the other, although they both attempt to characterize the same phenomena. We should ideally start from an integrated description of both the model and the steps carried out to test it, to concurrently analyze uncertainties in model parameters, equipment tolerances, and data collection. To this end, we present a language to model and optimize experimental biochemical protocols that facilitates such an integrated description, and that can be combined with experimental data. We provide a probabilistic semantics for our language based on a Bayesian interpretation that formally characterizes the uncertainties in both the data collection, the underlying model, and the protocol operations. On a set of case studies we illustrate how the resulting framework allows for automated analysis and optimization of experimental protocols, including Gibson assembly protocols.

q-bio.QM↗

arXiv2021

Probabilistic Model Checking and Autonomy

Design and control of autonomous systems that operate in uncertain or adversarial environments can be facilitated by formal modelling and analysis. Probabilistic model checking is a technique to automatically verify, for a given temporal logic specification, that a system model satisfies the specification, as well as to synthesise an optimal strategy for its control. This method has recently been extended to multi-agent systems that exhibit competitive or cooperative behaviour modelled via stochastic games and synthesis of equilibria strategies. In this paper, we provide an overview of probabilistic model checking, focusing on models supported by the PRISM and PRISM-games model checkers. This includes fully observable and partially observable Markov decision processes, as well as turn-based and concurrent stochastic games, together with associated probabilistic temporal logics. We demonstrate the applicability of the framework through illustrative examples from autonomous systems. Finally, we highlight research challenges and suggest directions for future work in this area.

cs.LO↗

arXiv2021

Certification of Iterative Predictions in Bayesian Neural Networks

We consider the problem of computing reach-avoid probabilities for iterative predictions made with Bayesian neural network (BNN) models. Specifically, we leverage bound propagation techniques and backward recursion to compute lower bounds for the probability that trajectories of the BNN model reach a given set of states while avoiding a set of unsafe states. We use the lower bounds in the context of control and reinforcement learning to provide safety certification for given control policies, as well as to synthesize control policies that improve the certification bounds. On a set of benchmarks, we demonstrate that our framework can be employed to certify policies over BNNs predictions for problems of more than $10$ dimensions, and to effectively synthesize policies that significantly increase the lower bound on the satisfaction probability.

cs.LG↗

arXiv2021

Provable Guarantees on the Robustness of Decision Rules to Causal Interventions

Robustness of decision rules to shifts in the data-generating process is crucial to the successful deployment of decision-making systems. Such shifts can be viewed as interventions on a causal graph, which capture (possibly hypothetical) changes in the data-generating process, whether due to natural reasons or by the action of an adversary. We consider causal Bayesian networks and formally define the interventional robustness problem, a novel model-based notion of robustness for decision functions that measures worst-case performance with respect to a set of interventions that denote changes to parameters and/or causal influences. By relying on a tractable representation of Bayesian networks as arithmetic circuits, we provide efficient algorithms for computing guaranteed upper and lower bounds on the interventional robustness probabilities. Experimental results demonstrate that the methods yield useful and interpretable bounds for a range of practical networks, paving the way towards provably causally robust decision-making systems.

cs.AI↗

arXiv2021

On Guaranteed Optimal Robust Explanations for NLP Models

We build on abduction-based explanations for ma-chine learning and develop a method for computing local explanations for neural network models in natural language processing (NLP). Our explanations comprise a subset of the words of the in-put text that satisfies two key features: optimality w.r.t. a user-defined cost function, such as the length of explanation, and robustness, in that they ensure prediction invariance for any bounded perturbation in the embedding space of the left out words. We present two solution algorithms, respectively based on implicit hitting sets and maximum universal subsets, introducing a number of algorithmic improvements to speed up convergence of hard instances. We show how our method can be con-figured with different perturbation sets in the em-bedded space and used to detect bias in predictions by enforcing include/exclude constraints on biased terms, as well as to enhance existing heuristic-based NLP explanation frameworks such as Anchors. We evaluate our framework on three widely used sentiment analysis tasks and texts of up to100words from SST, Twitter and IMDB datasets,demonstrating the effectiveness of the derived explanations.

cs.AI↗

arXiv2021

Adversarial Robustness Guarantees for Gaussian Processes

Gaussian processes (GPs) enable principled computation of model uncertainty, making them attractive for safety-critical applications. Such scenarios demand that GP decisions are not only accurate, but also robust to perturbations. In this paper we present a framework to analyse adversarial robustness of GPs, defined as invariance of the model's decision to bounded perturbations. Given a compact subset of the input space $T\subseteq \mathbb{R}^d$, a point $x^*$ and a GP, we provide provable guarantees of adversarial robustness of the GP by computing lower and upper bounds on its prediction range in $T$. We develop a branch-and-bound scheme to refine the bounds and show, for any $ε> 0$, that our algorithm is guaranteed to converge to values $ε$-close to the actual values in finitely many iterations. The algorithm is anytime and can handle both regression and classification tasks, with analytical formulation for most kernels used in practice. We evaluate our methods on a collection of synthetic and standard benchmark datasets, including SPAM, MNIST and FashionMNIST. We study the effect of approximate inference techniques on robustness and demonstrate how our method can be used for interpretability. Our empirical results suggest that the adversarial robustness of GPs increases with accurate posterior estimation.

cs.LG↗

arXiv2021

Bayesian Inference with Certifiable Adversarial Robustness

We consider adversarial training of deep neural networks through the lens of Bayesian learning, and present a principled framework for adversarial training of Bayesian Neural Networks (BNNs) with certifiable guarantees. We rely on techniques from constraint relaxation of non-convex optimisation problems and modify the standard cross-entropy error model to enforce posterior robustness to worst-case perturbations in $ε$-balls around input points. We illustrate how the resulting framework can be combined with methods commonly employed for approximate inference of BNNs. In an empirical investigation, we demonstrate that the presented approach enables training of certifiably robust models on MNIST, FashionMNIST and CIFAR-10 and can also be beneficial for uncertainty calibration. Our method is the first to directly train certifiable BNNs, thus facilitating their deployment in safety-critical applications.

cs.LG↗

arXiv2020

Assessing Robustness of Text Classification through Maximal Safe Radius Computation

Neural network NLP models are vulnerable to small modifications of the input that maintain the original meaning but result in a different prediction. In this paper, we focus on robustness of text classification against word substitutions, aiming to provide guarantees that the model prediction does not change if a word is replaced with a plausible alternative, such as a synonym. As a measure of robustness, we adopt the notion of the maximal safe radius for a given input text, which is the minimum distance in the embedding space to the decision boundary. Since computing the exact maximal safe radius is not feasible in practice, we instead approximate it by computing a lower and upper bound. For the upper bound computation, we employ Monte Carlo Tree Search in conjunction with syntactic filtering to analyse the effect of single and multiple word substitutions. The lower bound computation is achieved through an adaptation of the linear bounding techniques implemented in tools CNN-Cert and POPQORN, respectively for convolutional and recurrent network models. We evaluate the methods on sentiment analysis and news classification models for four datasets (IMDB, SST, AG News and NEWS) and a range of embeddings, and provide an analysis of robustness trends. We also apply our framework to interpretability analysis and compare it with LIME.

cs.CL↗

arXiv2020

Safety Guarantees for Planning Based on Iterative Gaussian Processes

Gaussian Processes (GPs) are widely employed in control and learning because of their principled treatment of uncertainty. However, tracking uncertainty for iterative, multi-step predictions in general leads to an analytically intractable problem. While approximation methods exist, they do not come with guarantees, making it difficult to estimate their reliability and to trust their predictions. In this work, we derive formal probability error bounds for iterative prediction and planning with GPs. Building on GP properties, we bound the probability that random trajectories lie in specific regions around the predicted values. Namely, given a tolerance $ε> 0 $, we compute regions around the predicted trajectory values, such that GP trajectories are guaranteed to lie inside them with probability at least $1-ε$. We verify experimentally that our method tracks the predictive uncertainty correctly, even when current approximation techniques fail. Furthermore, we show how the proposed bounds can be employed within a safe reinforcement learning framework to verify the safety of candidate control policies, guiding the synthesis of provably safe controllers.

cs.LG↗

arXiv2020

Automatic Verification of Concurrent Stochastic Systems

Automated verification techniques for stochastic games allow formal reasoning about systems that feature competitive or collaborative behaviour among rational agents in uncertain or probabilistic settings. Existing tools and techniques focus on turn-based games, where each state of the game is controlled by a single player, and on zero-sum properties, where two players or coalitions have directly opposing objectives. In this paper, we present automated verification techniques for concurrent stochastic games (CSGs), which provide a more natural model of concurrent decision making and interaction. We also consider (social welfare) Nash equilibria, to formally identify scenarios where two players or coalitions with distinct goals can collaborate to optimise their joint performance. We propose an extension of the temporal logic rPATL for specifying quantitative properties in this setting and present corresponding algorithms for verification and strategy synthesis for a variant of stopping games. For finite-horizon properties the computation is exact, while for infinite-horizon it is approximate using value iteration. For zero-sum properties it requires solving matrix games via linear programming, and for equilibria-based properties we find social welfare or social cost Nash equilibria of bimatrix games via the method of labelled polytopes through an SMT encoding. We implement this approach in PRISM-games, which required extending the tool's modelling language for CSGs, and apply it to case studies from domains including robotics, computer security and computer networks, explicitly demonstrating the benefits of both CSGs and equilibria-based properties.

cs.LO↗

arXiv2020

Multi-player Equilibria Verification for Concurrent Stochastic Games

Concurrent stochastic games (CSGs) are an ideal formalism for modelling probabilistic systems that feature multiple players or components with distinct objectives making concurrent, rational decisions. Examples include communication or security protocols and multi-robot navigation. Verification methods for CSGs exist but are limited to scenarios where agents or players are grouped into two coalitions, with those in the same coalition sharing an identical objective. In this paper, we propose multi-coalitional verification techniques for CSGs. We use subgame-perfect social welfare (or social cost) optimal Nash equilibria, which are strategies where there is no incentive for any coalition to unilaterally change its strategy in any game state, and where the total combined objectives are maximised (or minimised). We present an extension of the temporal logic rPATL (probabilistic alternating-time temporal logic with rewards) to specify equilibria-based properties for any number of distinct coalitions, and a corresponding model checking algorithm for a variant of stopping games. We implement our techniques in the PRISM-games tool and apply them to several case studies, including a secret sharing protocol and a public good game.

cs.LO↗

arXiv2020

Probabilistic Safety for Bayesian Neural Networks

We study probabilistic safety for Bayesian Neural Networks (BNNs) under adversarial input perturbations. Given a compact set of input points, $T \subseteq \mathbb{R}^m$, we study the probability w.r.t. the BNN posterior that all the points in $T$ are mapped to the same region $S$ in the output space. In particular, this can be used to evaluate the probability that a network sampled from the BNN is vulnerable to adversarial attacks. We rely on relaxation techniques from non-convex optimization to develop a method for computing a lower bound on probabilistic safety for BNNs, deriving explicit procedures for the case of interval and linear function propagation techniques. We apply our methods to BNNs trained on a regression task, airborne collision avoidance, and MNIST, empirically showing that our approach allows one to certify probabilistic safety of BNNs with millions of parameters.

cs.LG↗

arXiv2020

Invariant Causal Prediction for Block MDPs

Generalization across environments is critical to the successful application of reinforcement learning algorithms to real-world challenges. In this paper, we consider the problem of learning abstractions that generalize in block MDPs, families of environments with a shared latent state space and dynamics structure over that latent space, but varying observations. We leverage tools from causal inference to propose a method of invariant prediction to learn model-irrelevance state abstractions (MISA) that generalize to novel observations in the multi-environment setting. We prove that for certain classes of environments, this approach outputs with high probability a state abstraction corresponding to the causal feature set with respect to the return. We further provide more general bounds on model error and generalization error in the multi-environment setting, in the process showing a connection between causal variable selection and the state abstraction framework for MDPs. We give empirical evidence that our methods work in both linear and nonlinear settings, attaining improved generalization over single- and multi-task baselines.

cs.LG↗

arXiv2020

On the Benefits of Invariance in Neural Networks

Many real world data analysis problems exhibit invariant structure, and models that take advantage of this structure have shown impressive empirical performance, particularly in deep learning. While the literature contains a variety of methods to incorporate invariance into models, theoretical understanding is poor and there is no way to assess when one method should be preferred over another. In this work, we analyze the benefits and limitations of two widely used approaches in deep learning in the presence of invariance: data augmentation and feature averaging. We prove that training with data augmentation leads to better estimates of risk and gradients thereof, and we provide a PAC-Bayes generalization bound for models trained with data augmentation. We also show that compared to data augmentation, feature averaging reduces generalization error when used with convex losses, and tightens PAC-Bayes bounds. We provide empirical support of these theoretical results, including a demonstration of why generalization may not improve by training with data augmentation: the `learned invariance' fails outside of the training distribution.

cs.LG↗

arXiv2020

Robustness Guarantees for Deep Neural Networks on Videos

The widespread adoption of deep learning models places demands on their robustness. In this paper, we consider the robustness of deep neural networks on videos, which comprise both the spatial features of individual frames extracted by a convolutional neural network and the temporal dynamics between adjacent frames captured by a recurrent neural network. To measure robustness, we study the maximum safe radius problem, which computes the minimum distance from the optical flow sequence obtained from a given input to that of an adversarial example in the neighbourhood of the input. We demonstrate that, under the assumption of Lipschitz continuity, the problem can be approximated using finite optimisation via discretising the optical flow space, and the approximation has provable guarantees. We then show that the finite optimisation problem can be solved by utilising a two-player turn-based game in a cooperative setting, where the first player selects the optical flows and the second player determines the dimensions to be manipulated in the chosen flow. We employ an anytime approach to solve the game, in the sense of approximating the value of the game by monotonically improving its upper and lower bounds. We exploit a gradient-based search algorithm to compute the upper bounds, and the admissible A* algorithm to update the lower bounds. Finally, we evaluate our framework on the UCF101 video dataset.

cs.CV↗

arXiv2020

Adversarial Robustness Guarantees for Classification with Gaussian Processes

We investigate adversarial robustness of Gaussian Process Classification (GPC) models. Given a compact subset of the input space $T\subseteq \mathbb{R}^d$ enclosing a test point $x^*$ and a GPC trained on a dataset $\mathcal{D}$, we aim to compute the minimum and the maximum classification probability for the GPC over all the points in $T$. In order to do so, we show how functions lower- and upper-bounding the GPC output in $T$ can be derived, and implement those in a branch and bound optimisation algorithm. For any error threshold $ε> 0$ selected a priori, we show that our algorithm is guaranteed to reach values $ε$-close to the actual values in finitely many iterations. We apply our method to investigate the robustness of GPC models on a 2D synthetic dataset, the SPAM dataset and a subset of the MNIST dataset, providing comparisons of different GPC training techniques, and show how our method can be used for interpretability analysis. Our empirical analysis suggests that GPC robustness increases with more accurate posterior estimation.

stat.ML↗