Autonomous agents that interact with an environment usually To alleviate this issue, one can consider methods that condition face tasks that comprise complex, entangled behaviors over the policy or the value function on the specification of long horizons. Conventional reinforcement learning (RL) the whole task (Schaul et al. 2015) and such approaches were methods have successfully addressed this. However, in cases recently also proposed for tasks with non-Markovian reward when the agent is meant to perform several tasks across similar functions (Vaezipoor et al. 2021). However, the methods that environments, training a policy for every task separately specify the whole task usually rely on a blackbox neural network can be time-consuming and requires a lot of data. In such for planning when determining which sub-goal to reach cases, the agent can utilize a method that has built-in generalization next. This makes it hard to interpret the plan to solve the task capabilities. One such method relies on the assumption and although they show promising results in practice, it is that reward functions of these tasks can be decomposed unclear whether and when these approaches will generalize into a linear combination of successor features (Barreto et al. to a new task.

Many modern reinforcement learning algorithms build on the actor-critic (AC) framework: iterative improvement of a policy (the actor) using policy improvement operators and iterative approximation of the policy's value (the critic). In contrast, the popular value-based algorithm family employs improvement operators in the value update, to iteratively improve the value function directly. In this work, we propose a general extension to the AC framework that employs two separate improvement operators: one applied to the policy in the spirit of policy-based algorithms and one applied to the value in the spirit of value-based algorithms, which we dub Value-Improved AC (VI-AC). We design two practical VI-AC algorithms based in the popular online off-policy AC algorithms TD3 and DDPG. We evaluate VI-TD3 and VI-DDPG in the Mujoco benchmark and find that both improve upon or match the performance of their respective baselines in all environments tested.

Designing effective reward functions is crucial to training reinforcement learning (RL) algorithms. However, this design is non-trivial, even for domain experts, due to the subjective nature of certain tasks that are hard to quantify explicitly. In recent works, large language models (LLMs) have been used for reward generation from natural language task descriptions, leveraging their extensive instruction tuning and commonsense understanding of human behavior. In this work, we hypothesize that LLMs, guided by human feedback, can be used to formulate human-aligned reward functions. Specifically, we study this in the challenging setting of autonomous driving (AD), wherein notions of "good" driving are tacit and hard to quantify. To this end, we introduce REvolve, an evolutionary framework that uses LLMs for reward design in AD. REvolve creates and refines reward functions by utilizing human feedback to guide the evolution process, effectively translating implicit human knowledge into explicit reward functions for training (deep) RL agents. We demonstrate that agents trained on REvolve-designed rewards align closely with human driving standards, thereby outperforming other state-of-the-art baselines.

Model-based offline Reinforcement Learning (RL) allows agents to fully utilise pre-collected datasets without requiring additional or unethical explorations. However, applying model-based offline RL to online systems presents challenges, primarily due to the highly suboptimal (noise-filled) and diverse nature of datasets generated by online systems. To tackle these issues, we introduce the Causal Prompting Reinforcement Learning (CPRL) framework, designed for highly suboptimal and resource-constrained online scenarios. The initial phase of CPRL involves the introduction of the Hidden-Parameter Block Causal Prompting Dynamic (Hip-BCPD) to model environmental dynamics. This approach utilises invariant causal prompts and aligns hidden parameters to generalise to new and diverse online users. In the subsequent phase, a single policy is trained to address multiple tasks through the amalgamation of reusable skills, circumventing the need for training from scratch. Experiments conducted across datasets with varying levels of noise, including simulation-based and real-world offline datasets from the Dnurse APP, demonstrate that our proposed method can make robust decisions in out-of-distribution and noisy environments, outperforming contemporary algorithms. Additionally, we separately verify the contributions of Hip-BCPDs and the skill-reuse strategy to the robustness of performance. We further analyse the visualised structure of Hip-BCPD and the interpretability of sub-skills. We released our source code and the first ever real-world medical dataset for precise medical decision-making tasks.

We introduce a novel framework of combinatorial multi-armed bandits (CMAB) with multivariant and probabilistically triggering arms (CMAB-MT), where the outcome of each arm is a $d$-dimensional multivariant random variable and the feedback follows a general arm triggering process. Compared with existing CMAB works, CMAB-MT not only enhances the modeling power but also allows improved results by leveraging distinct statistical properties for multivariant random variables. For CMAB-MT, we propose a general 1-norm multivariant and triggering probability-modulated smoothness condition, and an optimistic CUCB-MT algorithm built upon this condition. Our framework can include many important problems as applications, such as episodic reinforcement learning (RL) and probabilistic maximum coverage for goods distribution, all of which meet the above smoothness condition and achieve matching or improved regret bounds compared to existing works. Through our new framework, we build the first connection between the episodic RL and CMAB literature, by offering a new angle to solve the episodic RL through the lens of CMAB, which may encourage more interactions between these two important directions.

This work presents a novel approach to neural architecture search (NAS) that aims to increase carbon efficiency for the model design process. The proposed framework CE-NAS addresses the key challenge of high carbon cost associated with NAS by exploring the carbon emission variations of energy and energy differences of different NAS algorithms. At the high level, CE-NAS leverages a reinforcement-learning agent to dynamically adjust GPU resources based on carbon intensity, predicted by a time-series transformer, to balance energy-efficient sampling and energy-intensive evaluation tasks. Furthermore, CE-NAS leverages a recently proposed multi-objective optimizer to effectively reduce the NAS search space. We demonstrate the efficacy of CE-NAS in lowering carbon emissions while achieving SOTA results for both NAS datasets and open-domain NAS tasks. For example, on the HW-NasBench dataset, CE-NAS reduces carbon emissions by up to 7.22X while maintaining a search efficiency comparable to vanilla NAS. For open-domain NAS tasks, CE-NAS achieves SOTA results with 97.35% top-1 accuracy on CIFAR-10 with only 1.68M parameters and a carbon consumption of 38.53 lbs of CO2. On ImageNet, our searched model achieves 80.6% top-1 accuracy with a 0.78 ms TensorRT latency using FP16 on NVIDIA V100, consuming only 909.86 lbs of CO2, making it comparable to other one-shot-based NAS baselines.

Inverse reinforcement learning (IRL) aims to infer a reward from expert demonstrations, motivated by the idea that the reward, rather than the policy, is the most succinct and transferable description of a task [Ng et al., 2000]. However, the reward corresponding to an optimal policy is not unique, making it unclear if an IRL-learned reward is transferable to new transition laws in the sense that its optimal policy aligns with the optimal policy corresponding to the expert's true reward. Past work has addressed this problem only under the assumption of full access to the expert's policy, guaranteeing transferability when learning from two experts with the same reward but different transition laws that satisfy a specific rank condition [Rolland et al., 2022]. In this work, we show that the conditions developed under full access to the expert's policy cannot guarantee transferability in the more practical scenario where we have access only to demonstrations of the expert. Instead of a binary rank condition, we propose principal angles as a more refined measure of similarity and dissimilarity between transition laws. Based on this, we then establish two key results: 1) a sufficient condition for transferability to any transition laws when learning from at least two experts with sufficiently different transition laws, and 2) a sufficient condition for transferability to local changes in the transition law when learning from a single expert. Furthermore, we also provide a probably approximately correct (PAC) algorithm and an end-to-end analysis for learning transferable rewards from demonstrations of multiple experts.

Reward learning involves inferring and shaping the underlying reward function from observed human demonstrations and feedback. Inverse reinforcement learning (IRL) and reinforcement learning from human feedback (RLHF, also known as preference-based reinforcement learning) are key methodologies in reward learning, applied in various sequential decision-making tasks such as games [1-3], robotics [4-6], and language models [7-11]. Particularly in the recent drastic development of large language models (LLMs), RLHF has played a crucial role in fine-tuning models to better align with human preferences [10]. However, despite the notable empirical success of these algorithms, a significant gap remains in the theoretical analysis of IRL and RLHF, limiting us to guarantee their reliability. This work aims to bridge this gap by proposing a novel theoretical framework for offline IRL and RLHF. IRL aims to infer a reward function that aligns with an expert behavior from demonstrations [12, 13]. Typical IRL algorithms employ a bi-level optimization framework within the context of maximum likelihood estimation (MLE). In this framework, the inner optimization evaluates the policy based on the current reward parameters, while the outer optimization updates these parameters to better match observed expert behavior. These algorithms have been extensively explored in the literature [14-18], and their convergence is studied in both online settings [17] and offline settings [18].

In recent years, significant attention has been directed towards learning average-reward Markov Decision Processes (MDPs). However, existing algorithms either suffer from sub-optimal regret guarantees or computational inefficiencies. In this paper, we present the first tractable algorithm with minimax optimal regret of $\widetilde{\mathrm{O}}(\sqrt{\mathrm{sp}(h^*) S A T})$, where $\mathrm{sp}(h^*)$ is the span of the optimal bias function $h^*$, $S \times A$ is the size of the state-action space and $T$ the number of learning steps. Remarkably, our algorithm does not require prior information on $\mathrm{sp}(h^*)$. Our algorithm relies on a novel subroutine, Projected Mitigated Extended Value Iteration (PMEVI), to compute bias-constrained optimal policies efficiently. This subroutine can be applied to various previous algorithms to improve regret bounds.

In various applications, the optimal policy in a strategic decision-making problem depends both on the environmental configuration and exogenous events. For these settings, we introduce Bilevel Optimization with Contextual Markov Decision Processes (BO-CMDP), a stochastic bilevel decision-making model, where the lower level consists of solving a contextual Markov Decision Process (CMDP). BO-CMDP can be viewed as a Stackelberg Game where the leader and a random context beyond the leader's control together decide the setup of (many) MDPs that (potentially multiple) followers best respond to. This framework extends beyond traditional bilevel optimization and finds relevance in diverse fields such as model design for MDPs, tax design, reward shaping and dynamic mechanism design. We propose a stochastic Hyper Policy Gradient Descent (HPGD) algorithm to solve BO-CMDP, and demonstrate its convergence. Notably, HPGD only utilizes observations of the followers' trajectories. Therefore, it allows followers to use any training procedure and the leader to be agnostic of the specific algorithm used, which aligns with various real-world scenarios. We further consider the setting when the leader can influence the training of followers and propose an accelerated algorithm. We empirically demonstrate the performance of our algorithm.