Repeated games consider a situation where multiple agents are motivated by their independent rewards throughout learning. In general, the dynamics of their learning become complex. Especially when their rewards compete with each other like zero-sum games, the dynamics often do not converge to their optimum, i.e., the Nash equilibrium. To tackle such complexity, many studies have understood various learning algorithms as dynamical systems and discovered qualitative insights among the algorithms. However, such studies have yet to handle multi-memory games (where agents can memorize actions they played in the past and choose their actions based on their memories), even though memorization plays a pivotal role in artificial intelligence and interpersonal relationship. This study extends two major learning algorithms in games, i.e., replicator dynamics and gradient ascent, into multi-memory games. Then, we prove their dynamics are identical. Furthermore, theoretically and experimentally, we clarify that the learning dynamics diverge from the Nash equilibrium in multi-memory zero-sum games and reach heteroclinic cycles (sojourn longer around the boundary of the strategy space), providing a fundamental advance in learning in games.

Embodied Conversational Agents are virtually embodied agents with a human-like appearance that are capable of autonomously communicating with people in a socially intelligent manner using multimodal behaviors (Lugrin [2021]). The field of research in ECAs has emerged as new interface between humans and machines. ECAs behaviors are often modeled from human communicative behaviors. They are endowed with the capacities to recognize and generate verbal and non-verbal cues (Lugrin [2021]), and are envisioned to support humans in their daily lives. Our work revolves around modeling multimodal data and learning the complex correlations between the different modalities employed in human communication. More specifically, the objective is to model the multimodal ECAs' behavior with their behavior style. Human behavior style is a socially meaningful clustering of features found within and across multiple modalities, specifically in linguistic (Campbell-Kibler et al. [2006]), spoken behavior such as the speaking style conveyed by speech prosody (Moon et al. [2022], Obin [2011]), and nonverbal behavior such as hand gestures and body posture (Obermeier et al. [2015], Wagner et al. [2014]).

In this work, we consider the problem of collaborative multi-user reinforcement learning. In this setting there are multiple users with the same state-action space and transition probabilities but with different rewards. Under the assumption that the reward matrix of the $N$ users has a low-rank structure -- a standard and practically successful assumption in the offline collaborative filtering setting -- the question is can we design algorithms with significantly lower sample complexity compared to the ones that learn the MDP individually for each user. Our main contribution is an algorithm which explores rewards collaboratively with $N$ user-specific MDPs and can learn rewards efficiently in two key settings: tabular MDPs and linear MDPs. When $N$ is large and the rank is constant, the sample complexity per MDP depends logarithmically over the size of the state-space, which represents an exponential reduction (in the state-space size) when compared to the standard ``non-collaborative'' algorithms.

Zero-shot human-AI coordination holds the promise of collaborating with humans without human data. Prevailing methods try to train the ego agent with a population of partners via self-play. However, these methods suffer from two problems: 1) The diversity of a population with finite partners is limited, thereby limiting the capacity of the trained ego agent to collaborate with a novel human; 2) Current methods only provide a common best response for every partner in the population, which may result in poor zero-shot coordination performance with a novel partner or humans. To address these issues, we first propose the policy ensemble method to increase the diversity of partners in the population, and then develop a context-aware method enabling the ego agent to analyze and identify the partner's potential policy primitives so that it can take different actions accordingly. In this way, the ego agent is able to learn more universal cooperative behaviors for collaborating with diverse partners. We conduct experiments on the Overcooked environment, and evaluate the zero-shot human-AI coordination performance of our method with both behavior-cloned human proxies and real humans. The results demonstrate that our method significantly increases the diversity of partners and enables ego agents to learn more diverse behaviors than baselines, thus achieving state-of-the-art performance in all scenarios. We also open-source a human-AI coordination study framework on the Overcooked for the convenience of future studies.

Active search, in applications like environment monitoring or disaster response missions, involves autonomous agents detecting targets in a search space using decision making algorithms that adapt to the history of their observations. Active search algorithms must contend with two types of uncertainty: detection uncertainty and location uncertainty. The more common approach in robotics is to focus on location uncertainty and remove detection uncertainty by thresholding the detection probability to zero or one. In contrast, it is common in the sparse signal processing literature to assume the target location is accurate and instead focus on the uncertainty of its detection. In this work, we first propose an inference method to jointly handle both target detection and location uncertainty. We then build a decision making algorithm on this inference method that uses Thompson sampling to enable decentralized multi-agent active search. We perform simulation experiments to show that our algorithms outperform competing baselines that only account for either target detection or location uncertainty. We finally demonstrate the real world transferability of our algorithms using a realistic simulation environment we created on the Unreal Engine 4 platform with an AirSim plugin.

The online knapsack problem is a classic problem in the field of online algorithms. Its canonical version asks how to pack items of different values and weights arriving online into a capacity-limited knapsack so as to maximize the total value of the admitted items. Although optimal competitive algorithms are known for this problem, they may be fundamentally unfair, i.e., individual items may be treated inequitably in different ways. Inspired by recent attention to fairness in online settings, we develop a natural and practically-relevant notion of time fairness for the online knapsack problem, and show that the existing optimal algorithms perform poorly under this metric. We propose a parameterized deterministic algorithm where the parameter precisely captures the Pareto-optimal trade-off between fairness and competitiveness. We show that randomization is theoretically powerful enough to be simultaneously competitive and fair; however, it does not work well in practice, using trace-driven experiments. To further improve the trade-off between fairness and competitiveness, we develop a fair, robust (competitive), and consistent learning-augmented algorithm with substantial performance improvement in trace-driven experiments.

In cooperative Multi-Agent Planning (MAP), a set of goals has to be achieved by a set of agents. Independently of whether they perform a pre-assignment of goals to agents or they directly search for a solution without any goal assignment, most previous works did not focus on a fair distribution/achievement of goals by agents. This paper adapts well-known fairness schemes to MAP, and introduces two novel approaches to generate cost-aware fair plans. The first one solves an optimization problem to pre-assign goals to agents, and then solves a centralized MAP task using that assignment. The second one consists of a planning-based compilation that allows solving the joint problem of goal assignment and planning while taking into account the given fairness scheme. Empirical results in several standard MAP benchmarks show that these approaches outperform different baselines. They also show that there is no need to sacrifice much plan cost to generate fair plans.

Evaluating grounded neural language model performance with respect to pragmatic qualities like the trade off between truthfulness, contrastivity and overinformativity of generated utterances remains a challenge in absence of data collected from humans. To enable such evaluation, we present a novel open source image-text dataset "Annotated 3D Shapes" (A3DS) comprising over nine million exhaustive natural language annotations and over 12 million variable-granularity captions for the 480,000 images provided by Burges & Kim (2018). We showcase the evaluation of pragmatic abilities developed by a task-neutral image captioner fine-tuned in a multi-agent communication setting to produce contrastive captions. The evaluation is enabled by the dataset because the exhaustive annotations allow to quantify the presence of contrastive features in the model's generations. We show that the model develops human-like patterns (informativity, brevity, over-informativity for specific features (e.g., shape, color biases)).

VoxML is a modeling language used to map natural language expressions into real-time visualizations using commonsense semantic knowledge of objects and events. Its utility has been demonstrated in embodied simulation environments and in agent-object interactions in situated multimodal human-agent collaboration and communication. It introduces the notion of object affordance (both Gibsonian and Telic) from HRI and robotics, as well as the concept of habitat (an object's context of use) for interactions between a rational agent and an object. This paper aims to specify VoxML as an annotation language in general abstract terms. It then shows how it works on annotating linguistic data that express visually perceptible human-object interactions. The annotation structures thus generated will be interpreted against the enriched minimal model created by VoxML as a modeling language while supporting the modeling purposes of VoxML linguistically.

In conventional distributed learning over a network, multiple agents collaboratively build a common machine learning model. However, due to the underlying non-i.i.d. data distribution among agents, the unified learning model becomes inefficient for each agent to process its locally accessible data. To address this problem, we propose a graph-attention-based personalized training algorithm (GATTA) for distributed deep learning. The GATTA enables each agent to train its local personalized model while exploiting its correlation with neighboring nodes and utilizing their useful information for aggregation. In particular, the personalized model in each agent is composed of a global part and a node-specific part. By treating each agent as one node in a graph and the node-specific parameters as its features, the benefits of the graph attention mechanism can be inherited. Namely, instead of aggregation based on averaging, it learns the specific weights for different neighboring nodes without requiring prior knowledge about the graph structure or the neighboring nodes' data distribution. Furthermore, relying on the weight-learning procedure, we develop a communication-efficient GATTA by skipping the transmission of information with small aggregation weights. Additionally, we theoretically analyze the convergence properties of GATTA for non-convex loss functions. Numerical results validate the excellent performances of the proposed algorithms in terms of convergence and communication cost.