论文检索

Reinforcement Learning · Deep RL

Chongyi Zheng, Jens Tuyls, Joanne Peng, Benjamin Eysenbach

Self-supervised learning has the potential of lifting several of the key challenges in reinforcement learning today, such as exploration, representation learning, and reward design. Recent work (METRA) has effectively argued that moving away from mutual information and instead optimizing a certain Wasserstein distance is important for good performance. In this paper, we argue that the benefits seen in that paper can largely be explained within the existing framework of mutual information skill learning (MISL).Our analysis suggests a new MISL method (contrastive successor features) that retains the excellent performance of METRA with fewer moving parts, and highlights connections between skill learning, contrastive representation learning, and successor features. Finally, through careful ablation studies, we provide further insight into some of the key ingredients for both our method and METRA.

Reinforcement Learning · Deep RL

Claas Voelcker, Marcel Hussing, ERIC EATON, Amir-massoud Farahmand, Igor Gilitschenski

Building deep reinforcement learning (RL) agents that find a good policy with few samples has proven notoriously challenging. To achieve sample efficiency, recent work has explored updating neural networks with large numbers of gradient steps for every new sample. While such high update-to-data (UTD) ratios have shown strong empirical performance, they also introduce instability to the training process. Previous approaches need to rely on periodic neural network parameter resets to address this instability, but restarting the training process is infeasible in many real-world applications and requires tuning the resetting interval. In this paper, we focus on one of the core difficulties of stable training with limited samples: the inability of learned value functions to generalize to unobserved on-policy actions. We mitigate this issue directly by augmenting the off-policy RL training process with a small amount of data generated from a learned world model. Our method, Model-Augmented Data for TD Learning (MAD-TD) uses small amounts of generated data to stabilize high UTD training and achieve competitive performance on the most challenging tasks in the DeepMind control suite. Our experiments further highlight the importance of employing a good model to generate data, MAD-TD's ability to combat value overestimation, and its practical stability gains for continued learning.

Yinan Zhang, Eric Tzeng, Yilun Du, Dmitry Kislyuk

Text-to-image diffusion models are a class of deep generative models that have demonstrated an impressive capacity for high-quality image generation. However, these models are susceptible to implicit biases that arise from web-scale text-image training pairs and may inaccurately model aspects of images we care about. This can result in suboptimal samples, model bias, and images that do not align with human ethics and preferences. In this paper, we present an effective scalable algorithm to improve diffusion models using Reinforcement Learning (RL) across a diverse set of reward functions, such as human preference, compositionality, and fairness over millions of images. We illustrate how our approach substantially outperforms existing methods for aligning diffusion models with human preferences. We further illustrate how this substantially improves pretrained Stable Diffusion (SD) models, generating samples that are preferred by humans 80.3% of the time over those from the base SD model while simultaneously improving both the composition and diversity of generated samples.

Jihyeong Jeon, Jiwon Park, Chanhee Park, U Kang

How can we leverage inherent frequency features of stock signals for effective portfolio optimization? Portfolio optimization in the domain of finance revolves around strategically allocating assets to maximize returns. Recent advancements highlight the efficacy of deep learning and reinforcement learning (RL) in capturing temporal asset patterns for portfolio optimization. However, previous methodologies focusing on time-domain often fail to detect sudden market shifts and abrupt events because their models are overly tailored to prevalent patterns, resulting in significant losses. In this paper, we propose FreQuant (Adaptive Portfolio Optimization via Multi-Frequency Quantitative Analysis), an effective deep RL framework for portfolio optimization that fully operates in the frequency domain, tackling the limitations of time domain-focused models. By bringing the analysis into the frequency domain with the Discrete Fourier Transform, our framework captures both prominent and subtle market frequencies, enhancing its adaptability and stability in response to market shifts. This approach allows FreQuant to adeptly identify primary asset patterns while also effectively responding to less common and abrupt market events, providing a more accurate and comprehensive asset representation. Empirical validation on diverse real-world trading datasets underscores the remarkable performance of FreQuant, showing its superiority in terms of profitability. Notably, FreQuant achieves up to 2.1x higher Annualized Rate of Return and 2.9x higher Portfolio Value than the best-performing competitors.

Heyang Gao, Zexu Sun, Hao Yang 0045, Xu Chen 0017

Causal discovery with observational and interventional data plays an important role in numerous fields. Due to the costly and potentially risky nature of intervention experiments, selecting informative interventions is critical in real-world situations. Several recent works introduce Bayesian active learning to select interventions that maximize the expected information gain about the underlying causal relationship at each optimization step. However, there are still some limitations within these methods: (1) Local optimality. With multiple intervention experiments, selecting optimal intervention myopically at each step may drop into the local optimal point. (2) Expensive time cost. Optimizing the most informative intervention at each step is time-consuming and not suitable for adaptive experiments with strict inference speed requirements. In this study, we propose a novel method called Reinforcement Learning-based Causal Bayesian Experimental Design (RL-CBED) to reduce the risk of local optimality and accelerate intervention selection inference. Specifically, we formulate the active causal discovery problem as a partially observable Markov decision process (POMDP). We design an information gain-based sparse reward function and then improve it to a dense reward function, providing fine-grained feedback to help the RL policy learn more quickly in complex environments. Moreover, we theoretically prove that the Q-function estimator can be learned using only trajectories sampled from the prior, which can significantly reduce the time cost of training process, enabling the real-world application of our method. Extensive experiments on both synthetic and real world-inspired semi-synthetic datasets demonstrate the effectiveness of our proposed method.

Felix Wiebe, Niccolò Turcato, Alberto Dalla Libera, Chi Zhang, Theo Vincent, Shubham Vyas, Giulio Giacomuzzo, Ruggero Carli, Diego Romeres, Akhil Sathuluri 等

As artificial intelligence gains new capabilities, it becomes important to evaluate it on real-world tasks. In particular, the fields of robotics and reinforcement learning (RL) are lacking in standardized benchmarking tasks on real hardware. To facilitate reproducibility and stimulate algorithmic advancements, we held an AI Olympics competition at IJCAI 2023 conference based on the double pendulum system in the RealAIGym project where the participants were asked to develop a controller for the swing up and stabilization task. This paper presents the methods and results from the top participating teams and provides insights into the real-world performance of RL algorithms with respect to a baseline time-varying LQR controller.

Shengming Wang, Xiaocai Zhang, Jing Li, Xiaoyang Wei, Hoong Chuin Lau, Bing Tian Dai, Binbin Huang, Zhe Xiao, Xiuju Fu, Zheng Qin

In recent years, there are trends toward cleaner port environments through enforcement by imposed legislation. Transit optimisation of fuel-based port service boats like harbour tugs has emerged as a critical task to reduce fuel consumption and carbon emission. In this paper, an innovative learning-based method, comprising a Reinforcement Learning (RL) model together with a fuel consumption prediction model, was proposed to formulate fuel-saving transit routes. Firstly, an ensemble model is established by combining a Long Short-Term Memory (LSTM) model with a Multilayer Perceptron (MLP) model, predicting fuel use based on tugboat movement and environment factors. Subsequently, an innovative RL based on Deep Deterministic Policy Gradient (DDPG) framework is developed considering the characteristics and obstructions of waterway in Singapore as well as the environmental factors to learn the optimal transit strategy that minimizes fuel consumption. We also demonstrate the efficacy of the solution to generate routes from origin to destination terminals, exhibiting significantly reduced fuel consumption in comparison to real-world transit scenarios.

Wanqi Xue, Bo An, Shuicheng Yan, Zhongwen Xu

The complexity of designing reward functions has been a major obstacle to the wide application of deep reinforcement learning (RL) techniques. Describing an agent's desired behaviors and properties can be difficult, even for experts. A new paradigm called reinforcement learning from human preferences (or preference-based RL) has emerged as a promising solution, in which reward functions are learned from human preference labels among behavior trajectories. However, existing methods for preference-based RL are limited by the need for accurate oracle preference labels. This paper addresses this limitation by developing a method for learning from diverse human preferences. The key idea is to stabilize reward learning through regularization and correction in a latent space. To ensure temporal consistency, a strong constraint is imposed on the reward model that forces its latent space to be close to a non-parameterized distribution. Additionally, a confidence-based reward model ensembling method is designed to generate more stable and reliable predictions. The proposed method is tested on a variety of tasks in DMcontrol and Meta-world and has shown consistent and significant improvements over existing preference-based RL algorithms when learning from diverse feedback, paving the way for real-world applications of RL methods.

Georgios Tzannetos, Parameswaran Kamalaruban, Adish Singla

Curriculum design for reinforcement learning (RL) can speed up an agent's learning process and help it learn to perform well on complex tasks. However, existing techniques typically require domain-specific hyperparameter tuning, involve expensive optimization procedures for task selection, or are suitable only for specific learning objectives. In this work, we consider curriculum design in contextual multi-task settings where the agent's final performance is measured w.r.t. a target distribution over complex tasks. We base our curriculum design on the Zone of Proximal Development concept, which has proven to be effective in accelerating the learning process of RL agents for uniform distribution over all tasks. We propose a novel curriculum, ProCuRL-Target, that effectively balances the need for selecting tasks that are not too difficult for the agent while progressing the agent's learning toward the target distribution via leveraging task correlations. We theoretically justify the task selection strategy of ProCuRL-Target by analyzing a simple learning setting with REINFORCE learner model. Our experimental results across various domains with challenging target task distributions affirm the effectiveness of our curriculum strategy over state-of-the-art baselines in accelerating the training process of deep RL agents.

Michael Schlechtinger, Damaris Kosack, Franz Krause, Heiko Paulheim

In the rapidly evolving landscape of eCommerce, Artificial Intelligence (AI) based pricing algorithms, particularly those utilizing Reinforcement Learning (RL), are becoming increasingly prevalent. This rise has led to an inextricable pricing situation with the potential for market collusion. Our research employs an experimental oligopoly model of repeated price competition, systematically varying the environment to cover scenarios from basic economic theory to subjective consumer demand preferences. We also introduce a novel demand framework that enables the implementation of various demand models, allowing for a weighted blending of different models. In contrast to existing research in this domain, we aim to investigate the strategies and emerging pricing patterns developed by the agents, which may lead to a collusive outcome. Furthermore, we investigate a scenario where agents cannot observe their competitors’ prices. Finally, we provide a comprehensive legal analysis across all scenarios. Our findings indicate that RL-based AI agents converge to a collusive state characterized by the charging of supracompetitive prices, without necessarily requiring inter-agent communication. Implementing alternative RL algorithms, altering the number of agents or simulation settings, and restricting the scope of the agents’ observation space does not significantly impact the collusive market outcome behavior.

Varun Ananth

Considering that the human brain is the most powerful, generalizable, and energy-efficient computer we know of, it makes the most sense to look to neuroscience for ideas regarding deep learning model improvements. I propose one such idea, augmenting a traditional Advantage-Actor-Critic (A2C) model with additional learning signals akin to those in the brain. Pursuing this direction of research should hopefully result in a new reinforcement learning (RL) control paradigm that can learn from fewer examples, train with greater stability, and possibly consume less energy.

Raffaele Galliera

This research explores optimizing communication tasks with (Multi-Agent) Reinforcement Learning (RL/MARL) in Point-to-Point and Group Communication (GC) networks. The study initially applied RL for Congestion Control in networks with dynamic link properties, yielding competitive results. Then, it focused on the challenge of effective message dissemination in GC networks, by framing a novel game-theoretic formulation and designing methods to solve the task based on MARL and Graph Convolution. Future research will deepen the exploration of MARL in GC. This will contribute to both academic knowledge and practical advancements in the next generation of communication protocols.

Fahmid Morshed Fahid, Jonathan Rowe, Yeojin Kim, Shashank Srivastava, James Lester

Pedagogical planners can provide adaptive support to students in narrative-centered learning environments by dynamically scaffolding student learning and tailoring problem scenarios. Reinforcement learning (RL) is frequently used for pedagogical planning in narrative-centered learning environments. However, RL-based pedagogical planning raises significant challenges due to the scarcity of data for training RL policies. Most prior work has relied on limited-size datasets and offline RL techniques for policy learning. Unfortunately, offline RL techniques do not support on-demand exploration and evaluation, which can adversely impact the quality of induced policies. To address the limitation of data scarcity and offline RL, we propose INSIGHT, an online RL framework for training data-driven pedagogical policies that optimize student learning in narrative-centered learning environments. The INSIGHT framework consists of three components: a narrative-centered learning environment simulator, a simulated student agent, and an RL-based pedagogical planner agent, which uses a reward metric that is associated with effective student learning processes. The framework enables the generation of synthetic data for on-demand exploration and evaluation of RL-based pedagogical planning. We have implemented INSIGHT with OpenAI Gym for a narrative-centered learning environment testbed with rule-based simulated student agents and a deep Q-learning-based pedagogical planner. Our results show that online deep RL algorithms can induce near-optimal pedagogical policies in the INSIGHT framework, while offline deep RL algorithms only find suboptimal policies even with large amounts of data.

Baturay Saglam, Furkan Mutlu, Dogan Cicek, Suleyman Kozat

A widely-studied deep reinforcement learning (RL) technique known as Prioritized Experience Replay (PER) allows agents to learn from transitions sampled with non-uniform probability proportional to their temporal-difference (TD) error. Although it has been shown that PER is one of the most crucial components for the overall performance of deep RL methods in discrete action domains, many empirical studies indicate that it considerably underperforms off-policy actor-critic algorithms. We theoretically show that actor networks cannot be effectively trained with transitions that have large TD errors. As a result, the approximate policy gradient computed under the Q-network diverges from the actual gradient computed under the optimal Q-function. Motivated by this, we introduce a novel experience replay sampling framework for actor-critic methods, which also regards issues with stability and recent findings behind the poor empirical performance of PER. The introduced algorithm suggests a new branch of improvements to PER and schedules effective and efficient training for both actor and critic networks. An extensive set of experiments verifies our theoretical findings, showing that our method outperforms competing approaches and achieves state-of-the-art results over the standard off-policy actor-critic algorithms.

Yifan Wang, Qining Zhang, Lei Ying, Chuan Zhou

Lung cancer remains the leading cause of cancer-related death worldwide, and early diagnosis of lung cancer is critical for improving the survival rate of patients. Performing annual low-dose computed tomography (LDCT) screening among high-risk populations is the primary approach for early diagnosis. However, after each screening, whether to continue monitoring (with follow-up screenings) or to order a biopsy for diagnosis remains a challenging decision to make. Continuing with follow-up screenings may lead to delayed diagnosis but ordering a biopsy without sufficient evidence incurs unnecessary risk and cost. In this paper, we tackle the problem by an optimal stopping approach. Our proposed algorithm, called EarlyStop-RL, utilizes the structure of the Snell envelope for optimal stopping, and model-free deep reinforcement learning for making diagnosis decisions. Through evaluating our algorithm on a commonly used clinical trial dataset (the National Lung Screening Trial), we demonstrate that EarlyStop-RL has the potential to greatly enhance risk assessment and early diagnosis of lung cancer, surpassing the performance of two widely adopted clinical models, namely the Lung-RADS and the Brock model.

Ruining Zhang, Haoran Han, Maolong Lv, Qisong Yang, Jian Cheng

Extensive utilization of deep reinforcement learning (DRL) policy networks in diverse continuous control tasks has raised questions regarding performance degradation in expansive state spaces where the input state norm is larger than that in the training environment. This paper aims to uncover the underlying factors contributing to such performance deterioration when dealing with expanded state spaces, using a novel analysis technique known as state division. In contrast to prior approaches that employ state division merely as a post-hoc explanatory tool, our methodology delves into the intrinsic characteristics of DRL policy networks. Specifically, we demonstrate that the expansion of state space induces the activation function $\tanh$ to exhibit saturability, resulting in the transformation of the state division boundary from nonlinear to linear. Our analysis centers on the paradigm of the double-integrator system, revealing that this gradual shift towards linearity imparts a control behavior reminiscent of bang-bang control. However, the inherent linearity of the division boundary prevents the attainment of an ideal bang-bang control, thereby introducing unavoidable overshooting. Our experimental investigations, employing diverse RL algorithms, establish that this performance phenomenon stems from inherent attributes of the DRL policy network, remaining consistent across various optimization algorithms.

Tianpei Yang, Heng You, Jianye Hao, Yan Zheng, Matthew E. Taylor

Transfer learning (TL) has shown great potential to improve Reinforcement Learning (RL) efficiency by leveraging prior knowledge in new tasks. However, much of the existing TL research focuses on transferring knowledge between tasks that share the same state-action spaces. Further, transfer from multiple source tasks that have different state-action spaces is more challenging and needs to be solved urgently to improve the generalization and practicality of the method in real-world scenarios. This paper proposes TURRET (Transfer Using gRaph neuRal nETworks), to utilize the generalization capabilities of Graph Neural Networks (GNNs) to facilitate efficient and effective multi-source policy transfer learning in the state-action mismatch setting. TURRET learns a semantic representation by accounting for the intrinsic property of the agent through GNNs, which leads to a unified state embedding space for all tasks. As a result, TURRET achieves more efficient transfer with strong generalization ability between different tasks and can be easily combined with existing Deep RL algorithms. Experimental results show that TURRET significantly outperforms other TL methods on multiple continuous action control tasks, successfully transferring across robots with different state-action spaces.

Di Wu, Yuling Jiao, Li Shen, Haizhao Yang, Xiliang Lu

Deep reinforcement learning (RL) has shown remarkable success in specific offline decision-making scenarios, yet its theoretical guarantees are still under development. Existing works on offline RL theory primarily emphasize a few trivial settings, such as linear MDP or general function approximation with strong assumptions and independent data, which lack guidance for practical use. The coupling of deep learning and Bellman residuals makes this problem challenging, in addition to the difficulty of data dependence. In this paper, we establish a non-asymptotic estimation error of pessimistic offline RL using general neural network approximation with C-mixing data regarding the structure of networks, the dimension of datasets, and the concentrability of data coverage, under mild assumptions. Our result shows that the estimation error consists of two parts: the first converges to zero at a desired rate on the sample size with partially controllable concentrability, and the second becomes negligible if the residual constraint is tight. This result demonstrates the explicit efficiency of deep adversarial offline RL frameworks. We utilize the empirical process tool for C-mixing sequences and the neural network approximation theory for the Holder class to achieve this. We also develop methods to bound the Bellman estimation error caused by function approximation with empirical Bellman constraint perturbations. Additionally, we present a result that lessens the curse of dimensionality using data with low intrinsic dimensionality and function classes with low complexity. Our estimation provides valuable insights into the development of deep offline RL and guidance for algorithm model design.

Sagar Patel, Sangeetha Abdu Jyothi, Nina Narodytska

We present CrystalBox, a novel, model-agnostic, posthoc explainability framework for Deep Reinforcement Learning (DRL) controllers in the large family of input-driven environments which includes computer systems. We combine the natural decomposability of reward functions in input-driven environments with the explanatory power of decomposed returns. We propose an efficient algorithm to generate future-based explanations across both discrete and continuous control environments. Using applications such as adaptive bitrate streaming and congestion control, we demonstrate CrystalBox's capability to generate high-fidelity explanations. We further illustrate its higher utility across three practical use cases: contrastive explanations, network observability, and guided reward design, as opposed to prior explainability techniques that identify salient features.

Jianfei Ma

Temporal difference learning (TD) is a foundational concept in reinforcement learning (RL), aimed at efficiently assessing a policy's value function. TD(λ), a potent variant, incorporates a memory trace to distribute the prediction error into the historical context. However, this approach often neglects the significance of historical states and the relative importance of propagating the TD error, influenced by challenges such as visitation imbalance or outcome noise. To address this, we propose a novel TD algorithm named discerning TD learning (DTD), which allows flexible emphasis functions—predetermined or adapted during training—to allocate efforts effectively across states. We establish the convergence properties of our method within a specific class of emphasis functions and showcase its promising potential for adaptation to deep RL contexts. Empirical results underscore that employing a judicious emphasis function not only improves value estimation but also expedites learning across diverse scenarios.