The availability of data is limited in some fields, especially for object detection tasks, where it is necessary to have correctly labeled bounding boxes around each object. A notable example of such data scarcity is found in the domain of marine biology, where it is useful to develop methods to automatically detect submarine species for environmental monitoring. To address this data limitation, the state-of-the-art machine learning strategies employ two main approaches. The first involves pretraining models on existing datasets before generalizing to the specific domain of interest. The second strategy is to create synthetic datasets specifically tailored to the target domain using methods like copy-paste techniques or ad-hoc simulators. The first strategy often faces a significant domain shift, while the second demands custom solutions crafted for the specific task. In response to these challenges, here we propose a transfer learning framework that is valid for a generic scenario. In this framework, generated images help to improve the performances of an object detector in a few-real data regime. This is achieved through a diffusion-based generative model that was pretrained on large generic datasets, and is not trained on the task-specific domain. We validate our approach on object detection tasks, specifically focusing on fishes in an underwater environment, and on the more common domain of cars in an urban setting. Our method achieves detection performance comparable to models trained on thousands of images, using only a few hundreds of input data. Our results pave the way for new generative AI-based protocols for machine learning applications in various domains, for instance ranging from geophysics to biology and medicine.

Simulation-to-reality (sim-to-real) transfer is a fundamental problem for robot learning. Domain Randomization, which adds randomization during training, is a powerful technique that effectively addresses the sim-to-real gap. However, the noise in observations makes learning significantly harder. Recently, studies have shown that employing a teacher-student learning paradigm can accelerate training in randomized environments. Learned with privileged information, a teacher agent can instruct the student agent to operate in noisy environments. However, this approach is often not sample efficient as the experience collected by the teacher is discarded completely when training the student, wasting information revealed by the environment. In this work, we extend the teacher-student learning paradigm by proposing a sample efficient learning framework termed Learn to Teach (L2T) that recycles experience collected by the teacher agent. We observe that the dynamics of the environments for both agents remain unchanged, and the state space of the teacher is coupled with the observation space of the student. We show that a single-loop algorithm can train both the teacher and student agents under both Reinforcement Learning and Inverse Reinforcement Learning contexts. We implement variants of our methods, conduct experiments on the MuJoCo benchmark, and apply our methods to the Cassie robot locomotion problem. Extensive experiments show that our method achieves competitive performance while only requiring environmental interaction with the teacher.

We develop optimal algorithms for learning undirected Gaussian trees and directed Gaussian polytrees from data. We consider both problems of distribution learning (i.e. in KL distance) and structure learning (i.e. exact recovery). The first approach is based on the Chow-Liu algorithm, and learns an optimal tree-structured distribution efficiently. The second approach is a modification of the PC algorithm for polytrees that uses partial correlation as a conditional independence tester for constraint-based structure learning. We derive explicit finite-sample guarantees for both approaches, and show that both approaches are optimal by deriving matching lower bounds. Additionally, we conduct numerical experiments to compare the performance of various algorithms, providing further insights and empirical evidence.

A recent theory shows that a multi-player decentralized partially observable Markov decision process can be transformed into an equivalent single-player game, enabling the application of \citeauthor{bellman}'s principle of optimality to solve the single-player game by breaking it down into single-stage subgames. However, this approach entangles the decision variables of all players at each single-stage subgame, resulting in backups with a double-exponential complexity. This paper demonstrates how to disentangle these decision variables while maintaining optimality under hierarchical information sharing, a prominent management style in our society. To achieve this, we apply the principle of optimality to solve any single-stage subgame by breaking it down further into smaller subgames, enabling us to make single-player decisions at a time. Our approach reveals that extensive-form games always exist with solutions to a single-stage subgame, significantly reducing time complexity. Our experimental results show that the algorithms leveraging these findings can scale up to much larger multi-player games without compromising optimality.

Recently, it has been shown that transformers pre-trained on diverse datasets with multi-episode contexts can generalize to new reinforcement learning tasks in-context. A key limitation of previously proposed models is their reliance on a predefined action space size and structure. The introduction of a new action space often requires data re-collection and model re-training, which can be costly for some applications. In our work, we show that it is possible to mitigate this issue by proposing the Headless-AD model that, despite being trained only once, is capable of generalizing to discrete action spaces of variable size, semantic content and order. By experimenting with Bernoulli and contextual bandits, as well as a gridworld environment, we show that Headless-AD exhibits significant capability to generalize to action spaces it has never encountered, even outperforming specialized models trained for a specific set of actions on several environment configurations.

Reinforcement learning (RL) is a classical tool to solve network control or policy optimization problems in unknown environments. The original Q-learning suffers from performance and complexity challenges across very large networks. Herein, a novel model-free ensemble reinforcement learning algorithm which adapts the classical Q-learning is proposed to handle these challenges for networks which admit Markov decision process (MDP) models. Multiple Q-learning algorithms are run on multiple, distinct, synthetically created and structurally related Markovian environments in parallel; the outputs are fused using an adaptive weighting mechanism based on the Jensen-Shannon divergence (JSD) to obtain an approximately optimal policy with low complexity. The theoretical justification of the algorithm, including the convergence of key statistics and Q-functions are provided. Numerical results across several network models show that the proposed algorithm can achieve up to 55% less average policy error with up to 50% less runtime complexity than the state-of-the-art Q-learning algorithms. Numerical results validate assumptions made in the theoretical analysis.

Motivated by the success of Transformers when applied to sequences of discrete symbols, token-based world models (TBWMs) were recently proposed as sample-efficient methods. In TBWMs, the world model consumes agent experience as a language-like sequence of tokens, where each observation constitutes a sub-sequence. However, during imagination, the sequential token-by-token generation of next observations results in a severe bottleneck, leading to long training times, poor GPU utilization, and limited representations. To resolve this bottleneck, we devise a novel Parallel Observation Prediction (POP) mechanism. POP augments a Retentive Network (RetNet) with a novel forward mode tailored to our reinforcement learning setting. We incorporate POP in a novel TBWM agent named REM (Retentive Environment Model), showcasing a 15.4x faster imagination compared to prior TBWMs. REM attains superhuman performance on 12 out of 26 games of the Atari 100K benchmark, while training in less than 12 hours. Our code is available at \url{https://github.com/leor-c/REM}.

The integration of physiological computing into mixed-initiative human-robot interaction systems offers valuable advantages in autonomous task allocation by incorporating real-time features as human state observations into the decision-making system. This approach may alleviate the cognitive load on human operators by intelligently allocating mission tasks between agents. Nevertheless, accommodating a diverse pool of human participants with varying physiological and behavioral measurements presents a substantial challenge. To address this, resorting to a probabilistic framework becomes necessary, given the inherent uncertainty and partial observability on the human's state. Recent research suggests to learn a Partially Observable Markov Decision Process (POMDP) model from a data set of previously collected experiences that can be solved using Offline Reinforcement Learning (ORL) methods. In the present work, we not only highlight the potential of partially observable representations and physiological measurements to improve human operator state estimation and performance, but also enhance the overall mission effectiveness of a human-robot team. Importantly, as the fixed data set may not contain enough information to fully represent complex stochastic processes, we propose a method to incorporate model uncertainty, thus enabling risk-sensitive sequential decision-making. Experiments were conducted with a group of twenty-six human participants within a simulated robot teleoperation environment, yielding empirical evidence of the method's efficacy. The obtained adaptive task allocation policy led to statistically significant higher scores than the one that was used to collect the data set, allowing for generalization across diverse participants also taking into account risk-sensitive metrics.

Infinite-state Markov Decision Processes (MDPs) are essential in modeling and optimizing a wide variety of engineering problems. In the reinforcement learning (RL) context, a variety of algorithms have been developed to learn and optimize these MDPs. At the heart of many popular policy-gradient based learning algorithms, such as natural actor-critic, TRPO, and PPO, lies the Natural Policy Gradient (NPG) algorithm. Convergence results for these RL algorithms rest on convergence results for the NPG algorithm. However, all existing results on the convergence of the NPG algorithm are limited to finite-state settings. We prove the first convergence rate bound for the NPG algorithm for infinite-state average-reward MDPs, proving a $O(1/\sqrt{T})$ convergence rate, if the NPG algorithm is initialized with a good initial policy. Moreover, we show that in the context of a large class of queueing MDPs, the MaxWeight policy suffices to satisfy our initial-policy requirement and achieve a $O(1/\sqrt{T})$ convergence rate. Key to our result are state-dependent bounds on the relative value function achieved by the iterate policies of the NPG algorithm.

We study inductive bias in Transformers in the infinitely over-parameterized Gaussian process limit and argue transformers tend to be biased towards more permutation symmetric functions in sequence space. We show that the representation theory of the symmetric group can be used to give quantitative analytical predictions when the dataset is symmetric to permutations between tokens. We present a simplified transformer block and solve the model at the limit, including accurate predictions for the learning curves and network outputs. We show that in common setups, one can derive tight bounds in the form of a scaling law for the learnability as a function of the context length. Finally, we argue WikiText dataset, does indeed possess a degree of permutation symmetry.