United States

We present a mean-variance policy iteration (MVPI) framework for risk-averse control in a discounted infinite horizon MDP optimizing the variance of a per-step reward random variable.
MVPI enjoys great flexibility in that
any policy evaluation method and risk-neutral control method can be dropped in for risk-averse control off the shelf,
in both on- and off-policy settings.
This flexibility reduces the gap between risk-neutral control and risk-averse control and is achieved by working on a novel augmented MDP directly.
We propose risk-averse TD3 as an example instantiating MVPI,
which outperforms vanilla TD3 and many previous risk-averse control methods in challenging Mujoco robot simulation tasks under a risk-aware performance metric.
This risk-averse TD3 is the first to introduce deterministic policies and off-policy learning into risk-averse reinforcement learning,
both of which are key to the performance boost we show in Mujoco domains.

AAAI 2021

Mean-Variance Policy Iteration for Risk-Averse Reinforcement Learning

We are pleased to announce the Thirty-Ninth AAAI Conference on Artificial Intelligence (AAAI-25), which will be held in Philadelphia, Pennsylvania at the Pennsylvania Convention Center from February 25 to March 4, 2025.

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-25 will feature technical paper presentations, special tracks, invited speakers, workshops, tutorials, poster sessions, senior member presentations, competitions, and exhibit programs, and a range of other activities to be announced.

### [Invited Speakers](https://aaai.org/conference/aaai/aaai-25/aaai-25-invited-speakers/)

Register [here](https://aaai.org/conference/aaai/aaai-25/registration/)

AAAI 2025

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-25 will feature technical paper presentations, special tracks, invited speakers, workshops, tutorials, poster sessions, senior member presentations, competitions, and exhibit programs, and a range of other activities to be announced.



technical paper

The purpose of the AAAI conference is to promote research in artificial intelligence (AI) and scientific exchange among AI researchers, practitioners, scientists, and engineers in affiliated disciplines. 

As an extension to the matroid span problem, we propose the submodular span problem that involves finding a large set of elements with small gain relative to a given query set. We then propose a two-stage Submodular Span Summarization (S3) framework to achieve a form of conditional or query-focused data summarization. The first stage encourages the summary to be relevant to a given query set, and the second stage encourages the final summary to be diverse, thus achieving two important necessities for a good query-focused summary. Unlike previous methods, our framework uses only a single submodular function defined over both data and query. We analyze theoretical properties in the context of both matroids and polymatroids that elucidate when our methods should work well. We find that a scalable approximation algorithm to the polymatroid submodular span problem has good theoretical and empirical properties. We provide empirical and qualitative results on three real-world tasks: conditional multi-document summarization on the DUC 2005-2007 datasets, conditional video summarization on the UT-Egocentric dataset, and conditional image corpus summarization on the ImageNet dataset. We use deep neural networks, specifically a BERT model for text, AlexNet for video frames, and Bi-directional Generative Adversarial Networks (BiGAN) for ImageNet images to help instantiate the submodular functions. The result is a minimally supervised form of conditional summarization that matches or improves over the previous state-of-the-art.

Submodular Span, with Applications to Conditional Data Summarization

Algorithms for noiseless collaborative PAC learning have been analyzed and optimized in recent years with respect to sample complexity. In this paper, we study collaborative PAC learning with the goal of reducing communication cost at essentially no penalty to the sample complexity. We develop communication efficient collaborative PAC learning algorithms using distributed boosting. We then consider the communication cost of collaborative learning in the presence of classification noise. As an intermediate step, we show how collaborative PAC learning algorithms can be adapted to handle classification noise. With this insight, we develop communication efficient algorithms for collaborative PAC learning robust to classification noise.

Communication-Aware Collaborative Learning

Current work in explainable reinforcement learning generally produces policies in the form of a decision tree over the state space. Such policies can be used for formal safety verification, agent behavior prediction, and manual inspection of important features. However, existing approaches fit a decision tree after training or use a custom learning procedure which is not compatible with new learning techniques, such as those which use neural networks. To address this limitation, we propose a novel Markov Decision Process (MDP) type for learning decision tree policies: Iterative Bounding MDPs (IBMDPs). An IBMDP is constructed around a base MDP so each IBMDP policy is guaranteed to correspond to a decision tree policy for the base MDP when using a method-agnostic masking procedure. Because of this decision tree equivalence, any function approximator can be used during training, including a neural network, while yielding a decision tree policy for the base MDP. We present the required masking procedure as well as a modified value update step which allows IBMDPs to be solved using existing algorithms. We apply this procedure to produce IBMDP variants of recent reinforcement learning methods. We empirically show the benefits of our approach by solving IBMDPs to produce decision tree policies for the base MDPs.

Iterative Bounding MDPs: Learning Interpretable Policies via Non-Interpretable Methods

This paper studies the adversarial graphical contextual bandits, a variant of adversarial multi-armed bandits that leverage two categories of the most common side information: contexts and side observations. In this setting, a learning agent repeatedly chooses from a set of K actions after being presented with a d-dimensional context vector. The agent not only incurs and observes the loss of the chosen action, but also observes the losses of its neighboring actions in the observation structures, which are encoded as a series of feedback graphs. This setting models a variety of applications in social networks, where both contexts and graph-structured side observations are available. Two efficient algorithms are developed based on EXP3. Under mild conditions, our analysis shows that for undirected feedback graphs the first algorithm, EXP3-LGC-U, achieves a sub-linear regret with respect to the time horizon and the average independence number of the feedback graphs. A slightly weaker result is presented for the directed graph setting as well. The second algorithm, EXP3-LGC-IX, is developed for a special class of problems, for which the regret is the same for both directed as well as undirected feedback graphs. Numerical tests corroborate the efficiency of proposed algorithms.

Adversarial Linear Contextual Bandits with Graph-Structured Side Observations

Models that top leaderboards often perform unsatisfactorily when deployed in real world applications; this has necessitated rigorous and expensive pre-deployment model testing. A hitherto unexplored facet of model performance is: Are our leaderboards doing equitable evaluation? In this paper, we introduce a task-agnostic method to probe leaderboards by weighting samples based on their 'difficulty' level. We find that leaderboards can be adversarially attacked and top performing models may not always be the best models. We subsequently propose alternate evaluation metrics. Our experiments on 10 models show changes in model ranking and an overall reduction in previously reported performance- thus rectifying the overestimation of AI systems' capabilities. Inspired by behavioral testing principles, we further develop a prototype of a visual analytics tool that enables leaderboard revamping through customization, based on an end user's focus area. This helps users analyze models' strengths and weaknesses, and guides them in the selection of a model best suited for their application scenario. In a user study, members of various commercial product development teams, covering 5 focus areas, find that our prototype reduces pre-deployment development and testing effort by 41% on average.

How Robust are Model Rankings : A Leaderboard Customization Approach for Equitable Evaluation

A split-transform-merge strategy has been broadly used as an architectural constraint in convolutional neural networks for visual recognition tasks. It approximates sparsely connected networks by explicitly defining multiple branches to simultaneously learn representations with different visual concepts or properties. Dependencies or interactions between these representations are typically defined by dense and local operations, however, without any adaptiveness or high-level reasoning. In this work, we propose to exploit this strategy and combine it with our Visual Concept Reasoning Networks (VCRNet) to enable reasoning between high-level visual concepts. We associate each branch with a visual concept and derive a compact concept state by selecting a few local descriptors through an attention module. These concept states are then updated by graph-based interaction and used to adaptively modulate the local descriptors. We describe our proposed model by split-transform-attend-interact-modulate-merge stages, which are implemented by opting for a highly modularized architecture. Extensive experiments on visual recognition tasks such as image classification, semantic segmentation, object detection, scene recognition, and action recognition show that our proposed model, VCRNet, consistently improves the performance by increasing the number of parameters by less than 1%.

Visual Concept Reasoning Networks

In this paper, we propose a normative approach to modeling apparently human irrational decision making (cognitive biases) that makes use of inherently rational computational mechanisms. We view preferential choice tasks as sequential decision making problems and formulate them as Partially Observable Markov Decision Processes (POMDPs). The resulting sequential decision model learns what information to gather about which options, whether to calculate option values or make comparisons between options and when to make a choice. We apply the model to choice problems where context is known to influence human choice, an effect that has been taken as evidence that human cognition is irrational. Our results show that the new model approximates a bounded optimal cognitive policy and makes quantitative predictions that correspond well to evidence about human choice. Furthermore, the model uses context to help infer which option has a maximum expected value while taking into account computational cost and cognitive limits. In addition, it predicts when, and explains why, people stop evidence accumulation and make a decision. We argue that the model provides evidence that apparent human irrationalities are emergent consequences of processes that prefer higher value (rational) policies.

Apparently Irrational Choice as Optimal Sequential Decision Making

Weight decay (WD) is a traditional regularization technique in deep learning, but despite its ubiquity, its behavior is still an area of active research. Golatkar et al. have recently shown that WD only matters at the start of the training in computer vision, upending traditional wisdom. Loshchilov et al. show that for adaptive optimizers, manually decaying weights can outperform adding an l2 penalty to the loss. This technique has become increasingly popular and is referred to as decoupled WD. The goal of this paper is to investigate these two recent empirical observations. We demonstrate that by applying WD only at the start, the network norm stays small throughout training. This has a regularizing effect as the effective gradient updates become larger. However, traditional generalizations metrics fail to capture this effect of WD, and we show how a simple scale-invariant metric can. We also show how the growth of network weights is heavily influenced by the dataset and its generalization properties. For decoupled WD, we perform experiments in NLP and RL where adaptive optimizers are the norm. We demonstrate that the primary issue that decoupled WD alleviates is the mixing of gradients from the objective function and the l2 penalty in the buffers of Adam (which stores the estimates of the first-order moment). Adaptivity itself is not problematic and decoupled WD ensures that the gradients from the l2 term cannot "drown out" the true objective, facilitating easier hyperparameter tuning.

Understanding Decoupled and Early Weight Decay

We propose a principled framework that combines adversarial training and provable robustness verification for training certifiably robust neural networks. We formulate the training problem as a joint optimization problem with both empirical and provable robustness objectives and develop a novel gradient-descent technique that can eliminate bias in stochastic multi-gradients. We perform both theoretical analysis on the convergence of the proposed technique and experimental comparison with state-of-the-arts. Results on MNIST and CIFAR-10 show that our method can consistently match or outperform prior approaches for provable lâˆž robustness. Notably, we achieve 6.60% verified test error on MNIST at Îµ = 0.3, and 66.57% on CIFAR-10 with Îµ = 8/255.

Adversarial Training and Provable Robustness: A Tale of Two Objectives

We consider the problem of minimizing a smooth, Lipschitz, convex function over a compact, convex set using sub-zeroth-order oracles: an oracle that outputs the sign of the directional derivative for a given point and a given direction, an oracle that compares the function values for a given pair of points, and an oracle that outputs a noisy function value for a given point. We show that the sample complexity of optimization using these oracles is polynomial in the relevant parameters. The optimization algorithm that we provide for the comparator oracle is the first algorithm with a known rate of convergence that is polynomial in the number of dimensions. We also give an algorithm for the noisy-value oracle that incurs sublinear regret in the number of queries and polynomial regret in the number of dimensions.

Downloads

Next from AAAI 2021

Submodular Span, with Applications to Conditional Data Summarization

Similar lecture

Risk-averse Total-reward MDPs with ERM and EVaR

.css-70qvj9{display:-webkit-box;display:-webkit-flex;display:-ms-flexbox;display:flex;-webkit-align-items:center;-webkit-box-align:center;-ms-flex-align:center;align-items:center;}Downloads

Next from AAAI 2021

Submodular Span, with Applications to Conditional Data Summarization

Similar lecture

Risk-averse Total-reward MDPs with ERM and EVaR

Downloads