United States

Safe exploration is a common problem in reinforcement learning (RL) that aims to prevent agents from making disastrous decisions while exploring their environment. A family of approaches to this problem assume domain knowledge in the form of a (partial) model of this environment to decide upon the safety of an action. A so-called shield forces the RL agent to select only safe actions. However, for adoption in various applications, one must look beyond enforcing safety and also ensure the applicability of RL with good performance. We extend the applicability of shields via tight integration with state-of-the-art deep RL, and provide an extensive, empirical study in challenging, sparse-reward environments under partial observability. We show that a carefully integrated shield ensures safety and can improve the convergence rate and final performance of RL agents. We furthermore show that a shield can be used to bootstrap state-of-the-art RL agents: they remain safe after initial learning in a shielded setting, allowing us to disable a potentially too conservative shield eventually.

AAAI 2023

Safe Reinforcement Learning via Shielding under Partial Observability

planning with markov models

sequential decision making

reinforcement learning algorithms

The purpose of the AAAI conference series is to promote research in Artificial Intelligence (AI) and foster scientific exchange between researchers, practitioners, scientists, students, and engineers across the entirety of AI and its affiliated disciplines. AAAI-23 is the Thirty-Seventh AAAI Conference on Artificial Intelligence. The theme of this conference is to create collaborative bridges within and beyond AI. Like previous AAAI conferences, AAAI-23 will feature technical paper presentations, special tracks, invited speakers, workshops, tutorials, poster sessions, senior member presentations, competitions, and exhibit programs, and two new activities: a Bridge Program and a Lab Program. Many of these activities are tailored to the theme of bridges and all are selected according to the highest standards, with additional programs for students and young researchers. 
AAAI is providing you with a conference planner, which you can use to help organize your itinerary of activities. This includes talks to attend in person, talks to attend remotely, breaks with colleagues and your site seeing activities. To access this conference planner, please go to [https://aaai-2023.takemobi.io](https://aaai-2023.takemobi.io).

In order to access this site, you need to register. If you haven't already, please register [here](https://aaai.org/Conferences/AAAI-23/registration/).


The purpose of the AAAI conference series is to promote research in Artificial Intelligence (AI) and foster scientific exchange between researchers, practitioners, scientists, students, and engineers across the entirety of AI and its affiliated disciplines.

poster

Steganography is a technique that hides secret messages into a public multimedia object without raising suspicion from third parties. However, most existing works cannot provide good robustness against lossy JPEG compression while maintaining a relatively large embedding capacity. This paper presents an end-to-end robust steganography system based on the invertible neural network (INN). Instead of hiding in the spatial domain, our method directly hides secret messages into the discrete cosine transform (DCT) coefficients of the cover image, which significantly improves the robustness and anti-steganalysis security. A mutual information loss is first proposed to constrain the flow of information in INN. Besides, a two-way fusion module (TWFM) is implemented, utilizing spatial and DCT domain features as auxiliary information to facilitate message extraction. These two designs aid in recovering secret messages from the DCT coefficients losslessly. Experimental results demonstrate that our method yields significantly lower error rates than other existing hiding methods. For example, our method achieves reliable extraction with 0 error rate for 1 bit per pixel (bpp) embedding payload; and under the JPEG compression with quality factor QF=10, the error rate of our method is about 22% lower than the state-of-the-art robust image hiding methods, which demonstrates remarkable robustness against JPEG compression.

Robust Image Steganography: Hiding Messages in Frequency Coefficients

Neural networks have complex structures, and thus it is hard to understand their inner workings and ensure correctness. To understand and debug convolutional neural networks (CNNs) we propose techniques for testing the channels of CNNs. We design FtGAN, an extension to GAN, that can generate test data with varying the intensity (i.e., sum of the neurons) of a channel of a target CNN. We also proposed a channel selection algorithm to find representative channels for testing. To efficiently inspect the target CNN’s inference computations, we define unexpectedness score, which estimates how similar the inference computation of the test data is to that of the training data. We evaluated FtGAN with five public datasets and showed that our techniques successfully identify defective channels in five different CNN models.

Testing the Channels of Convolutional Neural Networks

Deep neural networks are in the limelight of machine learning with their excellent performance in many data-driven applications. However, they can lead to inaccurate predictions when queried in out-of-distribution data points, which can have detrimental effects especially in sensitive domains, such as healthcare and transportation, where erroneous predictions can be very costly and/or dangerous. Subsequently, quantifying the uncertainty of the output of a neural network is often leveraged to evaluate the confidence of its predictions, and ensemble models have proved to be effective in measuring the uncertainty by utilizing the variance of predictions over a pool of models. In this paper, we propose a novel approach for uncertainty quantification via ensembles, called Random Activation Functions (RAFs) Ensemble, that aims at improving the ensemble diversity toward a more robust estimation, by accommodating each neural network with a different (random) activation function. Extensive empirical study demonstrates that RAFs Ensemble outperforms state-of-the-art ensemble uncertainty quantification methods on both synthetic and real-world datasets in a series of regression tasks.

Toward Robust Uncertainty Estimation with Random Activation Functions

There is a longstanding interest in capturing the error behaviour of object detectors by finding images where their performance is likely to be unsatisfactory. In real-world applications such as autonomous driving, it is also crucial to characterise potential failures beyond simple requirements of detection performance. For example, a missed detection of a pedestrian close to an ego vehicle will generally require closer inspection than a missed detection of a car in the distance. The problem of predicting such potential failures at test time  has largely been overlooked in the literature and conventional approaches based on detection uncertainty fall short in that they are agnostic to such fine-grained characterisation of errors. In this work, we propose to reformulate the problem of finding "hard" images as a query-based hard image retrieval task, where queries are specific definitions of "hardness", and offer a simple and intuitive method that can solve this task for a large family of queries. Our method is entirely post-hoc, does not require ground-truth annotations, is independent of the choice of a detector, and relies on an efficient Monte Carlo estimation that uses a simple stochastic model in place of the ground-truth. We show experimentally that it can be applied successfully to a wide variety of queries for which it can reliably identify hard images for a given detector without any labelled data. We provide results on ranking and classification tasks using the widely used RetinaNet, Faster-RCNN, Mask-RCNN, and Cascade Mask-RCNN object detectors.

Query-based Hard-Image Retrieval for Object Detection at Test Time

Communicating the predictive uncertainty of deep neural networks transparently and reliably is important in many safety-critical applications such as medicine. However, modern neural networks tend to be poorly calibrated, resulting in wrong
predictions made with a high confidence. While existing post-hoc calibration methods like temperature scaling or isotonic regression yield strongly calibrated predictions in artificial experimental settings, their efficiency can significantly reduce in real-world applications, where scarcity of labeled data or domain drifts are commonly present. In this paper, we first investigate the impact of these characteristics on post-hoc calibration and introduce an easy-to-implement extension of common post-hoc calibration methods based on test time augmentation. In extensive experiments, we demonstrate that our approach results in substantially better calibration on various architectures. We demonstrate the robustness of our proposed approach on a real-world application for skin cancer classification and show that it facilitates safe decision-making under real-world uncertainties.

Test Time Augmentation Meets Post-hoc Calibration: Uncertainty Quantification under Real-World Conditions

Graph neural networks (GNN) based collaborative filtering (CF) has attracted increasing attention in e-commerce and financial marketing platforms. However, there still lack efforts to evaluate the robustness of such CF systems in deployment. Fundamentally different from existing attacks, this work revisits the item promotion task and reformulates it from a targeted topological attack perspective for the first time. Specifically, we first develop a targeted attack formulation to maximally increase a target item's popularity. We then leverage gradient-based optimizations to find a solution. However, we observe the gradient estimates often appear noisy due to the discrete nature of a graph, which leads to a degradation of attack ability. To resolve noisy gradient effects, we then propose a masked attack objective that can remarkably enhance the topological attack ability. Furthermore, we design a computationally efficient approach to the proposed attack, thus making it feasible to evaluate large-large CF systems. Experiments on two real-world datasets show the effectiveness of our attack in analyzing the robustness of GNN-based CF more practically.

Revisiting Item Promotion in GNN-based Collaborative Filtering: A Masked Targeted Topological Attack Perspective

Safety is a critical hurdle that limits the application of deep reinforcement learning to real-world control tasks. To this end, constrained reinforcement learning leverages cost functions to improve safety in constrained Markov decision process. However, constrained methods fail to achieve zero violation even when the cost limit is zero. This paper analyzes the reason for such failure, which suggests that a proper cost function plays an important role in constrained RL. Inspired by the analysis, we propose AutoCost, a simple yet effective framework that automatically searches for cost functions that help constrained RL to achieve zero-violation performance. We validate the proposed method and the searched cost function on the safety benchmark Safety Gym. We compare the performance of augmented agents that use our cost function to provide additive intrinsic costs to a Lagrangian-based policy learner and a constrained-optimization policy learner with baseline agents that use the same policy learners but with only extrinsic costs. Results show that the converged policies with intrinsic costs in all environments achieve zero constraint violation and comparable performance with baselines.

AutoCost: Evolving Intrinsic Cost for Zero-violation Reinforcement Learning

This paper makes two key contributions. First, it argues that highly specialized rare content classifiers trained on small data typically have limited exposure to the richness and topical diversity of the negative class (dubbed \emph{anti-content}) as observed in the wild. As a result, these classifiers' strong performance observed on the test set may not translate into real-world settings. In the context of COVID-19 misinformation detection, we conduct an in-the-wild audit of multiple datasets and demonstrate that models trained with several prominently cited recent datasets are vulnerable to anti-content when evaluated in the wild. Second, we present a novel active learning pipeline that requires zero manual annotation and iteratively augments the training data with challenging anti-content, robustifying these classifiers. 

Auditing and Robustifying COVID-19 Misinformation Data Sets

In the scope of ``AI for Science'', solving inverse problems is a longstanding challenge in materials and drug discovery, where the goal is to determine the hidden structures given a set of desirable properties. Deep generative models are recently proposed to solve inverse problems, but these are currently struggling in expensive forward operators, precisely localizing the exact solutions and fully exploring the parameter spaces without missing solutions. In this work, we propose a novel approach (called {iPage}) to accelerate the inverse learning process by leveraging probabilistic inference from deep invertible models and deterministic optimization via fast gradient descent.  Given a target property, the learned invertible model provides a posterior over the parameter space; we identify these posterior samples as an intelligent prior initialization which enables us to narrow down the search space. We then perform gradient descent to calibrate the inverse solutions within a local region. Meanwhile, a space-filling sampling is imposed on the latent space to better explore and capture all possible solutions. We evaluate our approach on three benchmark tasks and create two datasets of real-world applications from quantum chemistry and additive manufacturing and find our method achieves superior performance compared to several state-of-the-art baseline methods. The iPage code is available at \url{https://github.com/jxzhangjhu/MatDesINNe}. 

Accelerating Inverse Learning via Intelligent Localization with Exploratory Sampling

Adversarial attacks have threatened modern deep learning systems by crafting adversarial examples with small perturbations to fool the convolutional neural networks (CNNs). To alleviate that, ensemble training methods are proposed to facilitate better adversarial robustness by diversifying the vulnerabilities among the sub-models, simultaneously maintaining comparable natural accuracy as standard training. Previous practices also demonstrate that enlarging the ensemble can improve the robustness. However, conventional ensemble methods are with poor scalability, owing to the rapidly increasing complexity when containing more sub-models in the ensemble. Moreover, it is usually infeasible to train or deploy an ensemble with substantial sub-models, owing to the tight hardware resource budget and time limits. In this work, we propose Ensemble-in-One (EIO), a simple but effective method to efficiently enlarge the ensemble with a random gated network (RGN). EIO augments a candidate model by replacing the parametrized layers with multi-path random gated blocks (RGBs) to construct an RGN. The scalability is significantly boosted because the number of paths exponentially increases with the RGN depth. Then by learning from the vulnerabilities of numerous other paths within the RGN, every path obtains better adversarial robustness. Our experiments demonstrate that EIO consistently outperforms previous ensemble training methods with smaller computational overheads, simultaneously achieving better accuracy-robustness trade-offs than adversarial training methods under black-box transfer attacks.

Downloads

Next from AAAI 2023

Robust Image Steganography: Hiding Messages in Frequency Coefficients

Similar lecture

Fast Counterfactual Inference for History-Based Reinforcement Learning