Deep Boltzmann machines (DBMs), one of the first ``deep'' learning methods ever studied, are multi-layered probabilistic models governed by a pairwise energy function that describes the likelihood of all variables/nodes in the network. In practice, DBMs are often constrained, i.e., via the \emph{restricted} Boltzmann machine (RBM) architecture (which does not permit intra-layer connections), in order to allow for more efficient inference. In this work, we revisit the generic DBM approach, and ask the question: are there other possible restrictions to their design that would enable efficient (approximate) inference? In particular, we develop a new class of restricted model, the monotone DBM, which allows for arbitrary self-connection in each layer, but restricts the \emph{weights} in a manner that guarantees the existence and global uniqueness of a mean-field fixed point. To do this, we leverage tools from the recently-proposed monotone Deep Equilibrium model and show that a particular choice of activation results in a fixed-point iteration that gives a variational mean-field solution. While this approach is still largely conceptual, it is the first architecture that allows for efficient approximate inference in fully-general weight structures for DBMs. We apply this approach to simple deep convolutional Boltzmann architectures and demonstrate that it allows for tasks such as the joint completion and classification of images, within a single deep probabilistic setting, while avoiding the pitfalls of mean-field inference in traditional RBMs.

In reinforcement learning, the objective is almost always defined as a \emph{cumulative} function over the rewards along the process. However, there are many optimal control and reinforcement learning problems in various application fields, especially in communications and networking, where the objectives are not naturally expressed as summations of the rewards. In this paper, we recognize the prevalence of non-cumulative objectives in various problems, and propose a modification to existing algorithms for optimizing such objectives. Specifically, we dive into the fundamental building block for many optimal control and reinforcement learning algorithms: the Bellman optimality equation. To optimize a non-cumulative objective, we replace the original summation operation in the Bellman update rule with a generalized operation corresponding to the objective. Furthermore, we provide sufficient conditions on the form of the generalized operation as well as assumptions on the Markov decision process under which the globally optimal convergence of the generalized Bellman updates can be guaranteed. We demonstrate the idea experimentally with the bottleneck objective, i.e., the objectives determined by the minimum reward along the process, on classical optimal control and reinforcement learning tasks, as well as on two network routing problems on maximizing the flow rates.

Probabilistic hyperproperties specify quantitative relations between the probabilities of reaching different target sets of states from different initial sets of states. This class of behavioral properties is suitable for capturing important security, privacy, and system-level requirements. We propose a new approach to solve the controller synthesis problem for Markov decision processes (MDPs) and probabilistic hyperproperties. Our specification language builds on top of the logic HyperPCTL and enhances it with structural constraints over the synthesized controllers. Our approach starts from a family of controllers represented symbolically and defined over the same copy of an MDP. We then introduce an abstraction refinement strategy that can relate multiple computation trees and that we employ to prune the search space deductively. The experimental evaluation demonstrates that the proposed approach considerably outperforms HyperProb, a state-of-the-art SMT-based model checking tool for HyperPCTL. Moreover, our approach is the first one that is able to effectively combine probabilistic hyperproperties with additional intra-controller constraints (e.g.

We provide time- and sample-efficient algorithms for learning and testing latent-tree Ising models, i.e. Ising models that may only be observed at their leaf nodes. On the learning side, we obtain efficient algorithms for learning a tree-structured Ising model whose leaf node distribution is close in Total Variation Distance, improving on the results of prior work. On the testing side, we provide an efficient algorithm with fewer samples for testing whether two latent-tree Ising models have leaf-node distributions that are close or far in Total Variation distance. We obtain our algorithms by showing novel localization results for the total variation distance between the leaf-node distributions of tree-structured Ising models, in terms of their marginals on pairs of leaves.

Robotic peg-in-hole assembly represents a critical area of investigation in robotic automation. The fusion of reinforcement learning (RL) and deep neural networks (DNNs) has yielded remarkable breakthroughs in this field. However, existing RL-based methods grapple with delivering optimal performance under the unique environmental and mission constraints of fusion applications. As a result, we propose an inventively designed RL-based approach. In contrast to alternative methods, our focus centers on enhancing the DNN architecture rather than the RL model. Our strategy receives and integrates data from the RGB camera and force/torque (F/T) sensor, training the agent to execute the peg-in-hole assembly task in a manner akin to human hand-eye coordination. All training and experimentation unfold within a realistic environment, and empirical outcomes demonstrate that this multi-sensor fusion approach excels in rigid peg-in-hole assembly tasks, surpassing the repeatable accuracy of the robotic arm utilized--0.1 mm--in uncertain and unstable conditions.

A major challenge in reinforcement learning is to develop practical, sample-efficient algorithms for exploration in high-dimensional domains where generalization and function approximation is required. Low-Rank Markov Decision Processes -- where transition probabilities admit a low-rank factorization based on an unknown feature embedding -- offer a simple, yet expressive framework for RL with function approximation, but existing algorithms are either (1) computationally intractable, or (2) reliant upon restrictive statistical assumptions such as latent variable structure, access to model-based function approximation, or reachability. In this work, we propose the first provably sample-efficient algorithm for exploration in Low-Rank MDPs that is both computationally efficient and model-free, allowing for general function approximation and requiring no additional structural assumptions. Our algorithm, VoX, uses the notion of a generalized optimal design for the feature embedding as an efficiently computable basis for exploration, performing efficient optimal design computation by interleaving representation learning and policy optimization. Our analysis -- which is appealingly simple and modular -- carefully combines several techniques, including a new reduction from optimal design computation to policy optimization based on the Frank-Wolfe method, and an improved analysis of a certain minimax representation learning objective found in prior work.

Text-based games provide a framework for developing natural language understanding and commonsense knowledge about the world in reinforcement learning based agents. Existing text-based environments often rely on fictional situations and characters to create a gaming framework and are far from real-world scenarios. In this paper, we introduce ScriptWorld: a text-based environment for teaching agents about real-world daily chores and hence imparting commonsense knowledge. To the best of our knowledge, it is the first interactive text-based gaming framework that consists of daily real-world human activities designed using scripts dataset. We provide gaming environments for 10 daily activities and perform a detailed analysis of the proposed environment. We develop RL-based baseline models/agents to play the games in Scriptworld. To understand the role of language models in such environments, we leverage features obtained from pre-trained language models in the RL agents. Our experiments show that prior knowledge obtained from a pre-trained language model helps to solve real-world text-based gaming environments. We release the environment via Github: https://github.com/Exploration-Lab/ScriptWorld

Systems and machines undergo various failure modes that result in machine health degradation, so maintenance actions are required to restore them back to a state where they can perform their expected functions. Since maintenance tasks are inevitable, maintenance planning is essential to ensure the smooth operations of the production system and other industries at large. Maintenance planning is a decision-making problem that aims at developing optimum maintenance policies and plans that help reduces maintenance costs, extend asset life, maximize their availability, and ultimately ensure workplace safety. Reinforcement learning is a data-driven decision-making algorithm that has been increasingly applied to develop dynamic maintenance plans while leveraging the continuous information from condition monitoring of the system and machine states. By leveraging the condition monitoring data of systems and machines with reinforcement learning, smart maintenance planners can be developed, which is a precursor to achieving a smart factory. This paper presents a literature review on the applications of reinforcement and deep reinforcement learning for maintenance planning and optimization problems. To capture the common ideas without losing touch with the uniqueness of each publication, taxonomies used to categorize the systems were developed, and reviewed publications were highlighted, classified, and summarized based on these taxonomies. Adopted methodologies, findings, and well-defined interpretations of the reviewed studies were summarized in graphical and tabular representations to maximize the utility of the work for both researchers and practitioners. This work also highlights the research gaps, key insights from the literature, and areas for future work.

Many different machine learning and deep learning techniques have been successfully employed for malware detection and classification. Examples of popular learning techniques in the malware domain include Hidden Markov Models (HMM), Random Forests (RF), Convolutional Neural Networks (CNN), Support Vector Machines (SVM), and Recurrent Neural Networks (RNN) such as Long Short-Term Memory (LSTM) networks. In this research, we consider a hybrid architecture, where HMMs are trained on opcode sequences, and the resulting hidden states of these trained HMMs are used as feature vectors in various classifiers. In this context, extracting the HMM hidden state sequences can be viewed as a form of feature engineering that is somewhat analogous to techniques that are commonly employed in Natural Language Processing (NLP). We find that this NLP-based approach outperforms other popular techniques on a challenging malware dataset, with an HMM-Random Forrest model yielding the best results.

Learning from rewards (i.e., reinforcement learning or RL) and learning to imitate a teacher (i.e., teacher-student learning) are two established approaches for solving sequential decision-making problems. To combine the benefits of these different forms of learning, it is common to train a policy to maximize a combination of reinforcement and teacher-student learning objectives. However, without a principled method to balance these objectives, prior work used heuristics and problem-specific hyperparameter searches to balance the two objectives. We present a $\textit{principled}$ approach, along with an approximate implementation for $\textit{dynamically}$ and $\textit{automatically}$ balancing when to follow the teacher and when to use rewards. The main idea is to adjust the importance of teacher supervision by comparing the agent's performance to the counterfactual scenario of the agent learning without teacher supervision and only from rewards. If using teacher supervision improves performance, the importance of teacher supervision is increased and otherwise it is decreased. Our method, $\textit{Teacher Guided Reinforcement Learning}$ (TGRL), outperforms strong baselines across diverse domains without hyper-parameter tuning.