Optimal and safety-critical control are fundamental problems for stochastic systems, and are widely considered in real-world scenarios such as robotic manipulation and autonomous driving. In this paper, we consider the problem of efficiently finding optimal and safe control for high-dimensional systems. Specifically, we propose to use dimensionality reduction techniques from a comparison theorem for stochastic differential equations together with a generalizable physics-informed neural network to estimate the optimal value function and the safety probability of the system. The proposed framework results in substantial sample efficiency improvement compared to existing methods. We further develop an autoencoder-like neural network to automatically identify the low-dimensional features in the system to enhance the ease of design for system integration. We also provide experiments and quantitative analysis to validate the efficacy of the proposed method. Source code is available at https://github.com/jacobwang925/path-integral-PINN.
论文检索
输入标题、作者或关键词，从 2,893 篇学术成果中精准定位
A plethora of fair graph neural networks (GNNs) have been proposed to promote algorithmic fairness for high-stake real-life contexts. Meanwhile, explainability is generally proposed to help machine learning practitioners debug models by providing human-understandable explanations. However, seldom work on explainability is made to generate explanations for fairness diagnosis in GNNs. From the explainability perspective, this paper explores the problem of what subgraph patterns cause the biased behavior of GNNs, and what actions could practitioners take to rectify the bias? By answering the two questions, this paper aims to produce compact, diagnostic, and actionable explanations that are responsible for discriminatory behavior. Specifically, we formulate the problem of generating diagnostic and actionable explanations as a multi-objective combinatorial optimization problem. To solve the problem, a dedicated multi-objective evolutionary algorithm is presented to ensure GNNs' explainability and fairness in one go. In particular, an influenced nodes-based gradient approximation is developed to boost the computation efficiency of the evolutionary algorithm. We provide a theoretical analysis to illustrate the effectiveness of the proposed framework. Extensive experiments have been conducted to demonstrate the superiority of the proposed method in terms of classification performance, fairness, and interpretability.
Warning: this paper contains data, prompts, and model outputs that are offensive in nature. Recently, there has been a surge of interest in integrating vision into Large Language Models (LLMs), exemplified by Visual Language Models (VLMs) such as Flamingo and GPT-4. This paper sheds light on the security and safety implications of this trend. First, we underscore that the continuous and high-dimensional nature of the visual input makes it a weak link against adversarial attacks, representing an expanded attack surface of vision-integrated LLMs. Second, we highlight that the versatility of LLMs also presents visual attackers with a wider array of achievable adversarial objectives, extending the implications of security failures beyond mere misclassification. As an illustration, we present a case study in which we exploit visual adversarial examples to circumvent the safety guardrail of aligned LLMs with integrated vision. Intriguingly, we discover that a single visual adversarial example can universally jailbreak an aligned LLM, compelling it to heed a wide range of harmful instructions (that it otherwise would not) and generate harmful content that transcends the narrow scope of a `few-shot' derogatory corpus initially employed to optimize the adversarial example. Our study underscores the escalating adversarial risks associated with the pursuit of multimodality. Our findings also connect the long-studied adversarial vulnerabilities of neural networks to the nascent field of AI alignment. The presented attack suggests a fundamental adversarial challenge for AI alignment, especially in light of the emerging trend toward multimodality in frontier foundation models.
Notions of transition invariants and closure certificates have seen recent use in the formal verification of controlled dynamical systems against \omega-regular properties. Unfortunately, existing approaches face limitations in two directions. First, they require a closed-form mathematical expression representing the model of the system. Such an expression may be difficult to find, too complex to be of any use, or unavailable due to security or privacy constraints. Second, finding such invariants typically rely on optimization techniques such as sum-of-squares (SOS) or satisfiability modulo theory (SMT) solvers. This restricts the classes of systems that need to be formally verified. To address these drawbacks, we introduce a notion of neural closure certificates. We present a data-driven algorithm that trains a neural network to represent a closure certificate. Our approach is formally correct under some mild assumptions, i.e., one is able to formally show that the unknown system satisfies the \omega-regular property of interest if a neural closure certificate can be computed. Finally, we demonstrate the efficacy of our approach with relevant case studies.
Unlearnable example attacks are data poisoning attacks aiming to degrade the clean test accuracy of deep learning by adding imperceptible perturbations to the training samples, which can be formulated as a bi-level optimization problem. However, directly solving this optimization problem is intractable for deep neural networks. In this paper, we investigate unlearnable example attacks from a game-theoretic perspective, by formulating the attack as a nonzero sum Stackelberg game. First, the existence of game equilibria is proved under the normal setting and the adversarial training setting. It is shown that the game equilibrium gives the most powerful poison attack in that the victim has the lowest test accuracy among all networks within the same hypothesis space when certain loss functions are used. Second, we propose a novel attack method, called the Game Unlearnable Example (GUE), which has three main gradients. (1) The poisons are obtained by directly solving the equilibrium of the Stackelberg game with a first-order algorithm. (2) We employ an autoencoder-like generative network model as the poison attacker. (3) A novel payoff function is introduced to evaluate the performance of the poison. Comprehensive experiments demonstrate that GUE can effectively poison the model in various scenarios. Furthermore, the GUE still works by using a relatively small percentage of the training data to train the generator, and the poison generator can generalize to unseen data well. Our implementation code can be found at https://github.com/hong-xian/gue.
Randomized smoothing is currently the state-of-the-art method that provides certified robustness for deep neural networks. However, due to its excessively conservative nature, this method of incomplete verification often cannot achieve an adequate certified radius on real-world datasets. One way to obtain a larger certified radius is to use an input-specific algorithm instead of using a fixed Gaussian filter for all data points. Several methods based on this idea have been proposed, but they either suffer from high computational costs or gain marginal improvement in certified radius. In this work, we show that by exploiting the quasiconvex problem structure, we can find the optimal certified radii for most data points with slight computational overhead. This observation leads to an efficient and effective input-specific randomized smoothing algorithm. We conduct extensive experiments and empirical analysis on CIFAR-10 and ImageNet. The results show that the proposed method significantly enhances the certified radii with low computational overhead.
There has been significant progress in improving the performance of graph neural networks (GNNs) through enhancements in graph data, model architecture design, and training strategies. For fairness in graphs, recent studies achieve fair representations and predictions through either graph data pre-processing (e.g., node feature masking, and topology rewiring) or fair training strategies (e.g., regularization, adversarial debiasing, and fair contrastive learning). How to achieve fairness in graphs from the model architecture perspective is less explored. More importantly, GNNs exhibit worse fairness performance compared to multilayer perception since their model architecture (i.e., neighbor aggregation) amplifies biases. To this end, we aim to achieve fairness via a new GNN architecture. We propose Fair Message Passing (FMP) designed within a unified optimization framework for GNNs. Notably, FMP explicitly renders sensitive attribute usage in forward propagation for node classification task using cross-entropy loss without data pre-processing. In FMP, the aggregation is first adopted to utilize neighbors' information and then the bias mitigation step explicitly pushes demographic group node presentation centers together. In this way, FMP scheme can aggregate useful information from neighbors and mitigate bias to achieve better fairness and prediction tradeoff performance. Experiments on node classification tasks demonstrate that the proposed FMP outperforms several baselines in terms of fairness and accuracy on three real-world datasets. The code is available at https://github.com/zhimengj0326/FMP.
π-Light: Programmatic Interpretable Reinforcement Learning for Resource-Limited Traffic Signal Control
PDF ↗The recent advancements in Deep Reinforcement Learning (DRL) have significantly enhanced the performance of adaptive Traffic Signal Control (TSC). However, DRL policies are typically represented by neural networks, which are over-parameterized black-box models. As a result, the learned policies often lack interpretability, and cannot be deployed directly in the real-world edge hardware due to resource constraints. In addition, the DRL methods often exhibit limited generalization performance, struggling to generalize the learned policy to other geographical regions. These factors limit the practical application of learning-based approaches. To address these issues, we suggest the use of an inherently interpretable program for representing the control policy. We present a new approach, Programmatic Interpretable reinforcement learning for traffic signal control (π-light), designed to autonomously discover non-differentiable programs. Specifically, we define a Domain Specific Language (DSL) and transformation rules for constructing programs, and utilize Monte Carlo Tree Search (MCTS) to find the optimal program in a discrete space. Extensive experiments demonstrate that our method consistently outperforms baseline approaches. Moreover, π-Light exhibits superior generalization capabilities compared to DRL, enabling training and evaluation across intersections from different cities. Finally, we analyze how the learned program policies can directly deploy on edge devices with extremely limited resources.
CASE: Exploiting Intra-class Compactness and Inter-class Separability of Feature Embeddings for Out-of-Distribution Detection
PDF ↗Detecting out-of-distribution (OOD) inputs is critical for reliable machine learning, but deep neural networks often make overconfident predictions, even for OOD inputs that deviate from the distribution of training data. Prior methods relied on the widely used softmax cross-entropy (CE) loss that is adequate for classifying in-distribution (ID) samples but not optimally designed for OOD detection. To address this issue, we propose CASE, a simple and effective OOD detection method by explicitly improving intra-class Compactness And inter-class Separability of feature Embeddings. To enhance the separation between ID and OOD samples, CASE uses a dual-loss framework, which includes a separability loss that maximizes the inter-class Euclidean distance to promote separability among different class centers, along with a compactness loss that minimizes the intra-class Euclidean distance to encourage samples to be close to their class centers. In particular, the class centers are defined as a free optimization parameter of the model and updated by gradient descent, which is simple and further enhances the OOD detection performance. Extensive experiments demonstrate the superiority of CASE, which reduces the average FPR95 by 37.11% and improves the average AUROC by 15.89% compared to the baseline method using a softmax confidence score on the more challenging CIFAR-100 model.
Neural-based, data-driven analysis and control of dynamical systems have been recently investigated and have shown great promise, e.g. for safety verification or stability analysis. Indeed, not only do neural networks allow for an entirely model-free, data-driven approach, but also for handling arbitrary complex functions via their power of representation (as opposed to, e.g. algebraic optimization techniques that are restricted to polynomial functions). Whilst classical Lyapunov techniques allow to provide a formal and robust guarantee of stability of a switched dynamical system, very little is yet known about correctness guarantees for Neural Lyapunov functions, nor about their performance (amount of data needed for a certain accuracy). We formally introduce Neural Lyapunov functions for the stability analysis of switched linear systems: we benchmark them on this paradigmatic problem, which is notoriously difficult (and in general Turing-undecidable), but which admits existing recently-developed technologies and theoretical results. Inspired by switched systems theory, we provide theoretical guarantees on the representative power of neural networks, leveraging recent results from the ML community. We additionally experimentally display how Neural Lyapunov functions compete with state-of-the-art results and techniques, while admitting a wide range of improvement, both in theory and in practice. This study intends to improve our understanding of the opportunities and current limitations of neural-based data-driven analysis and control of complex dynamical systems.
Improving Neural Network Generalization on Data-Limited Regression with Doubly-Robust Boosting
PDF ↗Enhancing the generalization performance of neural networks given limited data availability remains a formidable challenge, due to the model selection trade-off between training error and generalization gap. To handle this challenge, we present a posterior optimization issue, specifically designed to reduce the generalization error of trained neural networks. To operationalize this concept, we propose a Doubly-Robust Boosting machine (DRBoost) which consists of a statistical learner and a zero-order optimizer. The statistical learner reduces the model capacity and thus the generalization gap; the zero-order optimizer minimizes the training error in a gradient-free manner. The two components cooperate to reduce the generalization error of a fully trained neural network in a doubly robust manner. Furthermore, the statistical learner alleviates the multicollinearity in the discriminative layer and enhances the generalization performance. The zero-order optimizer eliminates the reliance on gradient calculation and offers more flexibility in learning objective selection. Experiments demonstrate that DRBoost improves the generalization performance of various prevalent neural network backbones effectively.
Combinatorial Optimization (CO) problems over graphs appear routinely in many applications such as in optimizing traffic, viral marketing in social networks, and matching for job allocation. Due to their combinatorial nature, these problems are often NP-hard. Existing approximation algorithms and heuristics rely on the search space to find the solutions and become time-consuming when this space is large. In this paper, we design a neural method called COMBHelper to reduce this space and thus improve the efficiency of the traditional CO algorithms based on node selection. Specifically, it employs a Graph Neural Network (GNN) to identify promising nodes for the solution set. This pruned search space is then fed to the traditional CO algorithms. COMBHelper also uses a Knowledge Distillation (KD) module and a problem-specific boosting module to bring further efficiency and efficacy. Our extensive experiments show that the traditional CO algorithms with COMBHelper are at least 2 times faster than their original versions.
Deep-reinforcement-learning (DRL) based neural combinatorial optimization (NCO) methods have demonstrated efficiency without relying on the guidance of optimal solutions. As the most mainstream among them, the learning constructive heuristic (LCH) achieves high-quality solutions through a rapid autoregressive solution construction process. However, these LCH-based methods are deficient in convergency, and there is still a performance gap compared to the optimal. Intuitively, learning to regret some steps in the solution construction process is helpful to the training efficiency and network representations. This article proposes a novel regret-based mechanism for an advanced solution construction process. Our method can be applied as a plug-in to any existing LCH-based DRL-NCO method. Experimental results demonstrate the capability of our work to enhance the performance of various NCO models. Results also show that the proposed LCH-Regret outperforms the previous modification methods on several typical combinatorial optimization problems. The code and Supplementary File are available at https://github.com/SunnyR7/LCH-Regret.
We present a fast, scalable, data-driven approach for solving relaxations of 0-1 integer linear programs. We use a combination of graph neural networks (GNN) and a Lagrange decomposition based algorithm. We make the latter differentiable for end-to-end training and use GNNs to predict its algorithmic parameters. This allows to retain the algorithm's theoretical properties including dual feasibility and guaranteed non-decrease in the lower bound while improving it via training. We overcome suboptimal fixed points of the basic solver by additional non-parametric GNN update steps maintaining dual feasibility. For training we use an unsupervised loss. We train on smaller problems and test on larger ones showing strong generalization performance with a GNN comprising only around 10k parameters. Our solver achieves significantly faster performance and better dual objectives than its non-learned version, achieving close to optimal objective values of LP relaxations of very large structured prediction problems and on selected combinatorial ones. In particular, we achieve better objective values than specialized approximate solvers for specific problem classes while retaining their efficiency. Our solver has better any-time performance over a large time period compared to a commercial solver.
Langevin dynamics (LD) is widely used for sampling from distributions and for optimization. In this work, we derive a closed-form expression for the expected loss of preconditioned LD near stationary points of the objective function. We use the fact that at the vicinity of such points, LD reduces to an Ornstein–Uhlenbeck process, which is amenable to convenient mathematical treatment. Our analysis reveals that when the preconditioning matrix satisfies a particular relation with respect to the noise covariance, LD's expected loss becomes proportional to the rank of the objective's Hessian. We illustrate the applicability of this result in the context of neural networks, where the Hessian rank has been shown to capture the complexity of the predictor function but is usually computationally hard to probe. Finally, we use our analysis to compare SGD-like and Adam-like preconditioners and identify the regimes under which each of them leads to a lower expected loss.
Deep neural network (DNN) models have been proven vulnerable to backdoor attacks. One trend of backdoor attacks is developing more invisible and dynamic triggers to make attacks stealthier. However, these invisible and dynamic triggers can be inadvertently mitigated by some widely used passive denoising operations, such as image compression, making the efforts under this trend questionable. Another trend is to exploit the full potential of backdoor attacks by proposing new triggering paradigms, such as hibernated or opportunistic backdoors. In line with these trends, our work investigates the first conditional backdoor attack, where the backdoor is activated by a specific condition rather than pre-defined triggers. Specifically, we take the JPEG compression as our condition and jointly optimize the compression operator and the target model's loss function, which can force the target model to accurately learn the JPEG compression behavior as the triggering condition. In this case, besides the conditional triggering feature, our attack is also stealthy and robust to denoising operations. Extensive experiments on the MNIST, GTSRB and CelebA verify our attack's effectiveness, stealthiness and resistance to existing backdoor defenses and denoising operations. As a new triggering paradigm, the conditional backdoor attack brings a new angle for assessing the vulnerability of DNN models, and conditioned over JPEG compression magnifies its threat due to the universal usage of JPEG.
Large language models (LLMs) and generative AI have played a transformative role in computer research and applications. Controversy has arisen as to whether these models output copyrighted data, which can occur if the data the models are trained on is copyrighted. LLMs are built on the transformer neural network architecture, which in turn relies on a mathematical computation called Attention that uses the softmax function. In this paper, we observe that large language model training and optimization can be seen as a softmax regression problem. We then establish a method of efficiently performing softmax regression, in a way that prevents the regression function from generating copyright data. This establishes a theoretical method of training large language models in a way that avoids generating copyright data.
Multi-agent Pathfinding (MAPF) problem generally asks to find a set of conflict-free paths for a set of agents confined to a graph and is typically solved in a centralized fashion. Conversely, in this work, we investigate the decentralized MAPF setting, when the central controller that possesses all the information on the agents' locations and goals is absent and the agents have to sequentially decide the actions on their own without having access to the full state of the environment. We focus on the practically important lifelong variant of MAPF, which involves continuously assigning new goals to the agents upon arrival to the previous ones. To address this complex problem, we propose a method that integrates two complementary approaches: planning with heuristic search and reinforcement learning through policy optimization. Planning is utilized to construct and re-plan individual paths. We enhance our planning algorithm with a dedicated technique tailored to avoid congestion and increase the throughput of the system. We employ reinforcement learning to discover the collision avoidance policies that effectively guide the agents along the paths. The policy is implemented as a neural network and is effectively trained without any reward-shaping or external guidance. We evaluate our method on a wide range of setups comparing it to the state-of-the-art solvers. The results show that our method consistently outperforms the learnable competitors, showing higher throughput and better ability to generalize to the maps that were unseen at the training stage. Moreover our solver outperforms a rule-based one in terms of throughput and is an order of magnitude faster than a state-of-the-art search-based solver. The code is available at https://github.com/AIRI-Institute/learn-to-follow.
ConcaveQ: Non-monotonic Value Function Factorization via Concave Representations in Deep Multi-Agent Reinforcement Learning
PDF ↗Value function factorization has achieved great success in multi-agent reinforcement learning by optimizing joint action-value functions through the maximization of factorized per-agent utilities. To ensure Individual-Global-Maximum property, existing works often focus on value factorization using monotonic functions, which are known to result in restricted representation expressiveness. In this paper, we analyze the limitations of monotonic factorization and present ConcaveQ, a novel non-monotonic value function factorization approach that goes beyond monotonic mixing functions and employs neural network representations of concave mixing functions. Leveraging the concave property in factorization, an iterative action selection scheme is developed to obtain optimal joint actions during training. It is used to update agents’ local policy networks, enabling fully decentralized execution. The effectiveness of the proposed ConcaveQ is validated across scenarios involving multi-agent predator-prey environment and StarCraft II micromanagement tasks. Empirical results exhibit significant improvement of ConcaveQ over state-of-the-art multi-agent reinforcement learning approaches.
Efficient communication learning among agents has been shown crucial for cooperative multi-agent reinforcement learning (MARL), as it can promote the action coordination of agents and ultimately improve performance. Graph neural network (GNN) provide a general paradigm for communication learning, which consider agents and communication channels as nodes and edges in a graph, with the action selection corresponding to node labeling. Under such paradigm, an agent aggregates information from neighbor agents, which can reduce uncertainty in local decision-making and induce implicit action coordination. However, this communication paradigm is vulnerable to adversarial attacks and noise, and how to learn robust and efficient communication under perturbations has largely not been studied. To this end, this paper introduces a novel Multi-Agent communication mechanism via Graph Information bottleneck (MAGI), which can optimally balance the robustness and expressiveness of the message representation learned by agents. This communication mechanism is aim at learning the minimal sufficient message representation for an agent by maximizing the mutual information (MI) between the message representation and the selected action, and simultaneously constraining the MI between the message representation and the agent feature. Empirical results demonstrate that MAGI is more robust and efficient than state-of-the-art GNN-based MARL methods.