Existing benchmarks for grounding language in interactive environments either lack real-world linguistic elements, or prove difficult to scale up due to substantial human involvement in the collection of data or feedback signals. To bridge this gap, we develop WebShop -- a simulated e-commerce website environment with $1.18$ million real-world products and $12,087$ crowd-sourced text instructions. Given a text instruction specifying a product requirement, an agent needs to navigate multiple types of webpages and issue diverse actions to find, customize, and purchase an item. WebShop provides several challenges for language grounding including understanding compositional instructions, query (re-)formulation, comprehending and acting on noisy text in webpages, and performing strategic exploration. We collect over $1,600$ human demonstrations for the task, and train and evaluate a diverse range of agents using reinforcement learning, imitation learning, and pre-trained image and language models. Our best model achieves a task success rate of $29\%$, which outperforms rule-based heuristics ($9.6\%$) but is far lower than human expert performance ($59\%$). We also analyze agent and human trajectories and ablate various model components to provide insights for developing future agents with stronger language understanding and decision making abilities. Finally, we show that agents trained on WebShop exhibit non-trivial sim-to-real transfer when evaluated on amazon.com and ebay.com, indicating the potential value of WebShop in developing practical web-based agents that can operate in the wild.

We introduce adversarial learning methods for data-driven generative modeling of the dynamics of $n^{th}$-order stochastic systems. Our approach builds on Generative Adversarial Networks (GANs) with generative model classes based on stable $m$-step stochastic numerical integrators. We introduce different formulations and training methods for learning models of stochastic dynamics based on observation of trajectory samples. We develop approaches using discriminators based on Maximum Mean Discrepancy (MMD), training protocols using conditional and marginal distributions, and methods for learning dynamic responses over different time-scales. We show how our approaches can be used for modeling physical systems to learn force-laws, damping coefficients, and noise-related parameters. The adversarial learning approaches provide methods for obtaining stable generative models for dynamic tasks including long-time prediction and developing simulations for stochastic systems.

Online damage quantification suffers from insufficient labeled data that weakens its accuracy. In this context, adopting the domain adaptation on historical labeled data from similar structures/damages or simulated digital twin data to assist the current diagnosis task would be beneficial. However, most domain adaptation methods are designed for classification and cannot efficiently address damage quantification, a regression problem with continuous real-valued labels. This study first proposes a novel domain adaptation method, the Online Fuzzy-set-based Joint Distribution Adaptation for Regression, to address this challenge. By converting the continuous real-valued labels to fuzzy class labels via fuzzy sets, the marginal and conditional distribution discrepancy are simultaneously measured to achieve the domain adaptation for the damage quantification task. Thanks to the superiority of the proposed method, a state-of-the-art online damage quantification framework based on domain adaptation is presented. Finally, the framework has been comprehensively demonstrated with a damaged helicopter panel, in which three types of damage domain adaptations (across different damage locations, across different damage types, and from simulation to experiment) are all conducted, proving the accuracy of damage quantification can be significantly improved in a realistic environment. It is expected that the proposed approach to be applied to the fleet-level digital twin considering the individual differences.

Existing episodic reinforcement algorithms assume that the length of an episode is fixed across time and known a priori. In this paper, we consider a general framework of episodic reinforcement learning when the length of each episode is drawn from a distribution. We first establish that this problem is equivalent to online reinforcement learning with general discounting where the learner is trying to optimize the expected discounted sum of rewards over an infinite horizon, but where the discounting function is not necessarily geometric. We show that minimizing regret with this new general discounting is equivalent to minimizing regret with uncertain episode lengths. We then design a reinforcement learning algorithm that minimizes regret with general discounting but acts for the setting with uncertain episode lengths. We instantiate our general bound for different types of discounting, including geometric and polynomial discounting. We also show that we can obtain similar regret bounds even when the uncertainty over the episode lengths is unknown, by estimating the unknown distribution over time. Finally, we compare our learning algorithms with existing value-iteration based episodic RL algorithms in a grid-world environment.

In this paper, we focus on the problem of efficiently locating a target object described with free-form language using a mobile robot equipped with vision sensors (e.g., an RGBD camera). Conventional active visual search predefines a set of objects to search for, rendering these techniques restrictive in practice. To provide added flexibility in active visual searching, we propose a system where a user can enter target commands using free-form language; we call this system Active Visual Search in the Wild (AVSW). AVSW detects and plans to search for a target object inputted by a user through a semantic grid map represented by static landmarks (e.g., desk or bed). For efficient planning of object search patterns, AVSW considers commonsense knowledge-based co-occurrence and predictive uncertainty while deciding which landmarks to visit first. We validate the proposed method with respect to SR (success rate) and SPL (success weighted by path length) in both simulated and real-world environments. The proposed method outperforms previous methods in terms of SPL in simulated scenarios with an average gap of 0.283. We further demonstrate AVSW with a Pioneer-3AT robot in real-world studies.

A key component of cognitive radar is the ability to generalize, or achieve consistent performance across a range of sensing environments, since aspects of the physical scene may vary over time. This presents a challenge for learning-based waveform selection approaches, since transmission policies which are effective in one scene may be highly suboptimal in another. We address this problem by strategically biasing a learning algorithm by exploiting high-level structure across tracking instances, referred to as meta-learning. In this work, we develop an online meta-learning approach for waveform-agile tracking. This approach uses information gained from previous target tracks to speed up and enhance learning in new tracking instances. This results in sample-efficient learning across a class of finite state target channels by exploiting inherent similarity across tracking scenes, attributed to common physical elements such as target type or clutter statistics. We formulate the online waveform selection problem within the framework of Bayesian learning, and provide prior-dependent performance bounds for the meta-learning problem using Probability Approximately Correct (PAC)-Bayes theory. We present a computationally feasible meta-posterior sampling algorithm and study the performance in a simulation study consisting of diverse scenes. Finally, we examine the potential performance benefits and practical challenges associated with online meta-learning for waveform-agile tracking.

Decentralized learning has shown great promise for cooperative multi-agent reinforcement learning (MARL). However, non-stationarity remains a significant challenge in fully decentralized learning. In the paper, we tackle the non-stationarity problem in the simplest and fundamental way and propose multi-agent alternate Q-learning (MA2QL), where agents take turns updating their Q-functions by Q-learning. MA2QL is a minimalist approach to fully decentralized cooperative MARL but is theoretically grounded. We prove that when each agent guarantees $\varepsilon$-convergence at each turn, their joint policy converges to a Nash equilibrium. In practice, MA2QL only requires minimal changes to independent Q-learning (IQL). We empirically evaluate MA2QL on a variety of cooperative multi-agent tasks. Results show MA2QL consistently outperforms IQL, which verifies the effectiveness of MA2QL, despite such minimal changes.

Given a machine learning (ML) model and a prediction, explanations can be defined as sets of features which are sufficient for the prediction. In some applications, and besides asking for an explanation, it is also critical to understand whether sensitive features can occur in some explanation, or whether a non-interesting feature must occur in all explanations. This paper starts by relating such queries respectively with the problems of relevancy and necessity in logic-based abduction. The paper then proves membership and hardness results for several families of ML classifiers. Afterwards the paper proposes concrete algorithms for two classes of classifiers. The experimental results confirm the scalability of the proposed algorithms.

The forward model, usually defined through partial differential equations (PDEs), describes certain physical phenomena with parameters as inputs. Generally, solving the forward model is computationally expensive but well-defined. In contrast, the related inverse problem which aims at inferring hidden parameters that cannot be directly observed from limited and noisy observations is typically ill-posed: different sets of parameters can result in similar sensor measurements, and there may be no feasible solution to fit the observed data, or minor errors can render unpredictable changes in the forward model. The Bayesian methods [5, 6], by viewing the unknown parameters as random variables, formulate the inverse problem into a probabilistic problem to capture the uncertainty in observations, forward models, and prior knowledge. One can assign a prior distribution to reflect our knowledge of the parameters before any measurements are made. The likelihood function is characterized through the forward model.

Among Bayesian methods, Monte-Carlo dropout provides principled tools for evaluating the epistemic uncertainty of neural networks. Its popularity recently led to seminal works that proposed activating the dropout layers only during inference for evaluating uncertainty. This approach, which we call dropout injection, provides clear benefits over its traditional counterpart (which we call embedded dropout) since it allows one to obtain a post hoc uncertainty measure for any existing network previously trained without dropout, avoiding an additional, time-consuming training process. Unfortunately, no previous work compared injected and embedded dropout; therefore, we provide the first thorough investigation, focusing on regression problems. The main contribution of our work is to provide guidelines on the effective use of injected dropout so that it can be a practical alternative to the current use of embedded dropout. In particular, we show that its effectiveness strongly relies on a suitable scaling of the corresponding uncertainty measure, and we discuss the trade-off between negative log-likelihood and calibration error as a function of the scale factor. Experimental results on UCI data sets and crowd counting benchmarks support our claim that dropout injection can effectively behave as a competitive post hoc uncertainty quantification technique.