PyTorchKR Papers

최신 논문

총 1,810편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:2503.17756

Bandwidth Reservation for Time-Critical Vehicular Applications: A Multi-Operator Environment

Onsite bandwidth reservation requests often face challenges such as price fluctuations and fairness issues due to unpredictable bandwidth availability and stringent latency requirements. Requesting bandwidth in advance can mitigate the impact of these fluctuations and ensure timely access to critical resources. In a multi-Mobile Network Operator (MNO) environment, vehicles need to select cost-effective and reliable resources for their safety-critical applications. This research aims to minimize resource costs by finding the best price among multiple MNOs. It formulates multi-operator scenarios as a Markov Decision Process (MDP), utilizing a Deep Reinforcement Learning (DRL) algorithm, specifically Dueling Deep Q-Learning. For efficient and stable learning, we propose a novel area-wise approach and an adaptive MDP synthetic close to the real environment. The Temporal Fusion Transformer (TFT) is used to handle time-dependent data and model training. Furthermore, the research leverages Amazon spot price data and adopts a multi-phase training approach, involving initial training on synthetic data, followed by real-world data. These phases enable the DRL agent to make informed decisions using insights from historical data and real-time observations. The results show that our model leads to significant cost reductions, up to 40%, compared to scenarios without a policy model in such a complex environment.

arXiv:2503.13557

APF+: Boosting adaptive-potential function reinforcement learning methods with a W-shaped network for high-dimensional games

Studies in reward shaping for reinforcement learning (RL) have flourished in recent years due to its ability to speed up training. Our previous work proposed an adaptive potential function (APF) and showed that APF can accelerate the Q-learning with a Multi-layer Perceptron algorithm in the low-dimensional domain. This paper proposes to extend APF with an encoder (APF+) for RL state representation, allowing applying APF to the pixel-based Atari games using a state-encoding method that projects high-dimensional game's pixel frames to low-dimensional embeddings. We approach by designing the state-representation encoder as a W-shaped network (W-Net), by using which we are able to encode both the background as well as the moving entities in the game frames. Specifically, the embeddings derived from the pre-trained W-Net consist of two latent vectors: One represents the input state, and the other represents the deviation of the input state's representation from itself. We then incorporate W-Net into APF to train a downstream Dueling Deep Q-Network (DDQN), obtain the APF-WNet-DDQN, and demonstrate its effectiveness in Atari game-playing tasks. To evaluate the APF+W-Net module in such high-dimensional tasks, we compare with two types of baseline methods: (i) the basic DDQN; and (ii) two encoder-replaced APF-DDQN methods where we replace W-Net by (a) an unsupervised state representation method called Spatiotemporal Deep Infomax (ST-DIM) and (b) a ground truth state representation provided by the Atari Annotated RAM Interface (ARI). The experiment results show that out of 20 Atari games, APF-WNet-DDQN outperforms DDQN (14/20 games) and APF-STDIM-DDQN (13/20 games) significantly. In comparison against the APF-ARI-DDQN which employs embeddings directly of the detailed game-internal state information, the APF-WNet-DDQN achieves a comparable performance.

arXiv:2503.11019

Residual Policy Gradient: A Reward View of KL-regularized Objective

Reinforcement Learning and Imitation Learning have achieved widespread success in many domains but remain constrained during real-world deployment. One of the main issues is the additional requirements that were not considered during training. To address this challenge, policy customization has been introduced, aiming to adapt a prior policy while preserving its inherent properties and meeting new task-specific requirements. A principled approach to policy customization is Residual Q-Learning (RQL), which formulates the problem as a Markov Decision Process (MDP) and derives a family of value-based learning algorithms. However, RQL has not yet been applied to policy gradient methods, which restricts its applicability, especially in tasks where policy gradient has already proven more effective. In this work, we first derive a concise form of Soft Policy Gradient as a preliminary. Building on this, we introduce Residual Policy Gradient (RPG), which extends RQL to policy gradient methods, allowing policy customization in gradient-based RL settings. With the view of RPG, we rethink the KL-regularized objective widely used in RL fine-tuning. We show that under certain assumptions, KL-regularized objective leads to a maximum-entropy policy that balances the inherent properties and task-specific requirements on a reward-level. Our experiments in MuJoCo demonstrate the effectiveness of Soft Policy Gradient and Residual Policy Gradient.

arXiv:2503.11270

Exploring Competitive and Collusive Behaviors in Algorithmic Pricing with Deep Reinforcement Learning

Nowadays, a significant share of the business-to-consumer sector is based on online platforms like Amazon and Alibaba and uses AI for pricing strategies. This has sparked debate on whether pricing algorithms may tacitly collude to set supra-competitive prices without being explicitly designed to do so. Our study addresses these concerns by examining the risk of collusion when Reinforcement Learning (RL) algorithms are used to decide on pricing strategies in competitive markets. Prior research in this field focused on Tabular Q-learning (TQL) and led to opposing views on whether learning-based algorithms can result in supra-competitive prices. Building on this, our work contributes to this ongoing discussion by providing a more nuanced numerical study that goes beyond TQL, additionally capturing off- and on- policy Deep Reinforcement Learning (DRL) algorithms, two distinct families of DRL algorithms that recently gained attention for algorithmic pricing. We study multiple Bertrand oligopoly variants and show that algorithmic collusion depends on the algorithm used. In our experiments, we observed that TQL tends to exhibit higher collusion and price dispersion. Moreover, it suffers from instability and disparity, as agents with higher learning rates consistently achieve higher profits, and it lacks robustness in state representation, with pricing dynamics varying significantly based on information access. In contrast, DRL algorithms, such as PPO and DQN, generally converge to lower prices closer to the Nash equilibrium. Additionally, we show that when pre-trained TQL agents interact with DRL agents, the latter quickly outperforms the former, highlighting the advantages of DRL in pricing competition. Lastly, we find that competition between heterogeneous DRL algorithms, such as PPO and DQN, tends to reduce the likelihood of supra-competitive pricing.

arXiv:2503.10186

Multi-Agent Q-Learning Dynamics in Random Networks: Convergence due to Exploration and Sparsity

Beyond specific settings, many multi-agent learning algorithms fail to converge to an equilibrium solution, and instead display complex, non-stationary behaviours such as recurrent or chaotic orbits. In fact, recent literature suggests that such complex behaviours are likely to occur when the number of agents increases. In this paper, we study Q-learning dynamics in network polymatrix games where the network structure is drawn from classical random graph models. In particular, we focus on the Erdos-Renyi model, a well-studied model for social networks, and the Stochastic Block model, which generalizes the above by accounting for community structures within the network. In each setting, we establish sufficient conditions under which the agents' joint strategies converge to a unique equilibrium. We investigate how this condition depends on the exploration rates, payoff matrices and, crucially, the sparsity of the network. Finally, we validate our theoretical findings through numerical simulations and demonstrate that convergence can be reliably achieved in many-agent systems, provided network sparsity is controlled.

arXiv:2503.09521공식 코드 · Papers with Code

PairVDN - Pair-wise Decomposed Value Functions

Extending deep Q-learning to cooperative multi-agent settings is challenging due to the exponential growth of the joint action space, the non-stationary environment, and the credit assignment problem. Value decomposition allows deep Q-learning to be applied at the joint agent level, at the cost of reduced expressivity. Building on past work in this direction, our paper proposes PairVDN, a novel method for decomposing the value function into a collection of pair-wise, rather than per-agent, functions, improving expressivity at the cost of requiring a more complex (but still efficient) dynamic programming maximisation algorithm. Our method enables the representation of value functions which cannot be expressed as a monotonic combination of per-agent functions, unlike past approaches such as VDN and QMIX. We implement a novel many-agent cooperative environment, Box Jump, and demonstrate improved performance over these baselines in this setting. We open-source our code and environment at https://github.com/zzbuzzard/PairVDN.

arXiv:2503.05970

Generative Multi-Agent Q-Learning for Policy Optimization: Decentralized Wireless Networks

Q-learning is a widely used reinforcement learning (RL) algorithm for optimizing wireless networks, but faces challenges with large state-spaces. Recently proposed multi-environment mixed Q-learning (MEMQ) algorithm addresses these challenges by employing multiple Q-learning algorithms across multiple synthetically generated, distinct but structurally related environments, so-called digital cousins. In this paper, we propose a novel multi-agent MEMQ (M-MEMQ) for cooperative decentralized wireless networks with multiple networked transmitters (TXs) and base stations (BSs). TXs do not have access to global information (joint state and actions). The new concept of coordinated and uncoordinated states is introduced. In uncoordinated states, TXs act independently to minimize their individual costs and update local Q-functions. In coordinated states, TXs use a Bayesian approach to estimate the joint state and update the joint Q-functions. The cost of information-sharing scales linearly with the number of TXs and is independent of the joint state-action space size. Several theoretical guarantees, including deterministic and probabilistic convergence, bounds on estimation error variance, and the probability of misdetecting the joint states, are given. Numerical simulations show that M-MEMQ outperforms several decentralized and centralized training with decentralized execution (CTDE) multi-agent RL algorithms by achieving 55% lower average policy error (APE), 35% faster convergence, 50% reduced runtime complexity, and 45% less sample complexity. Furthermore, M-MEMQ achieves comparable APE with significantly lower complexity than centralized methods. Simulations validate the theoretical analyses.

arXiv:2503.04219

Quantum-Inspired Reinforcement Learning in the Presence of Epistemic Ambivalence

The complexity of online decision-making under uncertainty stems from the requirement of finding a balance between exploiting known strategies and exploring new possibilities. Naturally, the uncertainty type plays a crucial role in developing decision-making strategies that manage complexity effectively. In this paper, we focus on a specific form of uncertainty known as epistemic ambivalence (EA), which emerges from conflicting pieces of evidence or contradictory experiences. It creates a delicate interplay between uncertainty and confidence, distinguishing it from epistemic uncertainty that typically diminishes with new information. Indeed, ambivalence can persist even after additional knowledge is acquired. To address this phenomenon, we propose a novel framework, called the epistemically ambivalent Markov decision process (EA-MDP), aiming to understand and control EA in decision-making processes. This framework incorporates the concept of a quantum state from the quantum mechanics formalism, and its core is to assess the probability and reward of every possible outcome. We calculate the reward function using quantum measurement techniques and prove the existence of an optimal policy and an optimal value function in the EA-MDP framework. We also propose the EA-epsilon-greedy Q-learning algorithm. To evaluate the impact of EA on decision-making and the expedience of our framework, we study two distinct experimental setups, namely the two-state problem and the lattice problem. Our results show that using our methods, the agent converges to the optimal policy in the presence of EA.

arXiv:2503.06741

A Novel Multi-Objective Reinforcement Learning Algorithm for Pursuit-Evasion Game

In practical application, the pursuit-evasion game (PEG) often involves multiple complex and conflicting objectives. The single-objective reinforcement learning (RL) usually focuses on a single optimization objective, and it is difficult to find the optimal balance among multiple objectives. This paper proposes a three-objective RL algorithm based on fuzzy Q-learning (FQL) to solve the PEG with different optimization objectives. First, the multi-objective FQL algorithm is introduced, which uses the reward function to represent three optimization objectives: evading pursuit, reaching target, and avoiding obstacle. Second, a multi-objective evaluation method and action selection strategy based on three-dimensional hypervolume are designed, which solved the dilemma of exploration-exploitation. By sampling the Pareto front, the update rule of the global strategy is obtained. The proposed algorithm reduces computational load while ensuring exploration ability. Finally, the performance of the algorithm is verified by simulation results.

arXiv:2503.04679

Multi-Agent Inverse Q-Learning from Demonstrations

When reward functions are hand-designed, deep reinforcement learning algorithms often suffer from reward misspecification, causing them to learn suboptimal policies in terms of the intended task objectives. In the single-agent case, inverse reinforcement learning (IRL) techniques attempt to address this issue by inferring the reward function from expert demonstrations. However, in multi-agent problems, misalignment between the learned and true objectives is exacerbated due to increased environment non-stationarity and variance that scales with multiple agents. As such, in multi-agent general-sum games, multi-agent IRL algorithms have difficulty balancing cooperative and competitive objectives. To address these issues, we propose Multi-Agent Marginal Q-Learning from Demonstrations (MAMQL), a novel sample-efficient framework for multi-agent IRL. For each agent, MAMQL learns a critic marginalized over the other agents' policies, allowing for a well-motivated use of Boltzmann policies in the multi-agent context. We identify a connection between optimal marginalized critics and single-agent soft-Q IRL, allowing us to apply a direct, simple optimization criterion from the single-agent domain. Across our experiments on three different simulated domains, MAMQL significantly outperforms previous multi-agent methods in average reward, sample efficiency, and reward recovery by often more than 2-5x. We make our code available at https://sites.google.com/view/mamql .

arXiv:2503.03515

DO-IQS: Dynamics-Aware Offline Inverse Q-Learning for Optimal Stopping with Unknown Gain Functions

We consider Inverse Optimal Stopping (IOS) problem where, based on stopped expert trajectories, one aims to recover the optimal stopping region through continuation and stopping gain functions approximation. The uniqueness of the stopping region allows the use of IOS in real-world applications with safety concerns. While current state-of-the-art inverse reinforcement learning methods recover both a Q-function and the corresponding optimal policy, they fail to account for specific challenges posed by optimal stopping problems. These include data sparsity near the stopping region, non-Markovian nature of the continuation gain, a proper treatment of boundary conditions, the need for a stable offline approach for risk-sensitive applications, and a lack of a quality evaluation metric. These challenges are addressed with the proposed Dynamics-Aware Offline Inverse Q-Learning for Optimal Stopping (DO-IQS), which incorporates temporal information by approximating the cumulative continuation gain together with the world dynamics and the Q-function without querying to the environment. Moreover, a confidence-based oversampling approach is proposed to treat the data sparsity problem. We demonstrate the performance of our models on real and artificial data including an optimal intervention for critical events problem.

arXiv:2503.00971

An Efficient and Uncertainty-aware Reinforcement Learning Framework for Quality Assurance in Extrusion Additive Manufacturing

Defects in extrusion additive manufacturing remain common despite its prevalent use. While numerous AI-driven approaches have been proposed to improve quality assurance, the inherently dynamic nature of the printing process poses persistent challenges. Regardless of how comprehensive the training dataset is, out-of-distribution data remains inevitable. Consequently, deterministic models often struggle to maintain robustness and, in some cases, fail entirely when deployed in new or slightly altered printing environments. This work introduces an agent that dynamically adjusts flow rate and temperature setpoints in real time, optimizing process control while addressing bottlenecks in training efficiency and uncertainty management. It integrates a vision-based uncertainty quantification module with a reinforcement learning controller, using probabilistic distributions to describe printing segments. While the underlying networks are deterministic, these evolving distributions introduce adaptability into the decision-making process. The vision system classifies material extrusion with a certain level of precision, generating corresponding distributions. A deep Q-learning controller interacts with a simulated environment calibrated to the vision system precision, allowing the agent to learn optimal actions while demonstrating appropriate hesitation when necessary. By executing asynchronous actions and applying progressively tightened elliptical reward shaping, the controller develops robust, adaptive control strategies that account for the coupling effects between process parameters. When deployed with zero-shot learning, the agent effectively bridges the sim-to-real gap, correcting mild and severe under- and over-extrusion reliably. Beyond extrusion additive manufacturing, this scalable framework enables practical AI-driven quality assurance across various additive manufacturing processes.

arXiv:2503.03338

Navigating Intelligence: A Survey of Google OR-Tools and Machine Learning for Global Path Planning in Autonomous Vehicles

We offer a new in-depth investigation of global path planning (GPP) for unmanned ground vehicles, an autonomous mining sampling robot named ROMIE. GPP is essential for ROMIE's optimal performance, which is translated into solving the traveling salesman problem, a complex graph theory challenge that is crucial for determining the most effective route to cover all sampling locations in a mining field. This problem is central to enhancing ROMIE's operational efficiency and competitiveness against human labor by optimizing cost and time. The primary aim of this research is to advance GPP by developing, evaluating, and improving a cost-efficient software and web application. We delve into an extensive comparison and analysis of Google operations research (OR)-Tools optimization algorithms. Our study is driven by the goal of applying and testing the limits of OR-Tools capabilities by integrating Reinforcement Learning techniques for the first time. This enables us to compare these methods with OR-Tools, assessing their computational effectiveness and real-world application efficiency. Our analysis seeks to provide insights into the effectiveness and practical application of each technique. Our findings indicate that Q-Learning stands out as the optimal strategy, demonstrating superior efficiency by deviating only 1.2% on average from the optimal solutions across our datasets.

arXiv:2503.00372

Nucleolus Credit Assignment for Effective Coalitions in Multi-agent Reinforcement Learning

In cooperative multi-agent reinforcement learning (MARL), agents typically form a single grand coalition based on credit assignment to tackle a composite task, often resulting in suboptimal performance. This paper proposed a nucleolus-based credit assignment grounded in cooperative game theory, enabling the autonomous partitioning of agents into multiple small coalitions that can effectively identify and complete subtasks within a larger composite task. Specifically, our designed nucleolus Q-learning could assign fair credits to each agent, and the nucleolus Q-operator provides theoretical guarantees with interpretability for both learning convergence and the stability of the formed small coalitions. Through experiments on Predator-Prey and StarCraft scenarios across varying difficulty levels, our approach demonstrated the emergence of multiple effective coalitions during MARL training, leading to faster learning and superior performance in terms of win rate and cumulative rewards especially in hard and super-hard environments, compared to four baseline methods. Our nucleolus-based credit assignment showed the promise for complex composite tasks requiring effective subteams of agents.

arXiv:2503.01450공식 코드 · Papers with Code

POPGym Arcade: Parallel Pixelated POMDPs

We present the POPGym Arcade, a collection of hardware-accelerated, pixel-based environments with shared observation and action spaces. Each environment includes fully and partially observable variants, enabling counterfactual studies on partial observability. We also introduce mathematical tools for analyzing policies under partial observability, which reveal how agents recall past information to make decisions. Our analysis shows (1) that controlling for partial observability is critical and (2) that agents with long-term memory learn brittle policies that struggle to generalize. Finally, we demonstrate that recurrent policies can be "poisoned" by old, out-of-distribution observations, with implications for sim-to-real transfer, imitation learning, and offline reinforcement learning.

arXiv:2412.12170

PickLLM: Context-Aware RL-Assisted Large Language Model Routing

Recently, the number of off-the-shelf Large Language Models (LLMs) has exploded with many open-source options. This creates a diverse landscape regarding both serving options (e.g., inference on local hardware vs remote LLM APIs) and model heterogeneous expertise. However, it is hard for the user to efficiently optimize considering operational cost (pricing structures, expensive LLMs-as-a-service for large querying volumes), efficiency, or even per-case specific measures such as response accuracy, bias, or toxicity. Also, existing LLM routing solutions focus mainly on cost reduction, with response accuracy optimizations relying on non-generalizable supervised training, and ensemble approaches necessitating output computation for every considered LLM candidate. In this work, we tackle the challenge of selecting the optimal LLM from a model pool for specific queries with customizable objectives. We propose PickLLM, a lightweight framework that relies on Reinforcement Learning (RL) to route on-the-fly queries to available models. We introduce a weighted reward function that considers per-query cost, inference latency, and model response accuracy by a customizable scoring function. Regarding the learning algorithms, we explore two alternatives: PickLLM router acting as a learning automaton that utilizes gradient ascent to select a specific LLM, or utilizing stateless Q-learning to explore the set of LLMs and perform selection with a $\epsilon$-greedy approach. The algorithm converges to a single LLM for the remaining session queries. To evaluate, we utilize a pool of four LLMs and benchmark prompt-response datasets with different contexts. A separate scoring function is assessing response accuracy during the experiment. We demonstrate the speed of convergence for different learning rates and improvement in hard metrics such as cost per querying session and overall response latency.

arXiv:2502.18984

Cycles and collusion in congestion games under Q-learning

We investigate the dynamics of Q-learning in a class of generalized Braess paradox games. These games represent an important class of network routing games where the associated stage-game Nash equilibria do not constitute social optima. We provide a full convergence analysis of Q-learning with varying parameters and learning rates. A wide range of phenomena emerges, broadly either settling into Nash or cycling continuously in ways reminiscent of "Edgeworth cycles" (i.e. jumping suddenly from Nash toward social optimum and then deteriorating gradually back to Nash). Our results reveal an important incentive incompatibility when thinking in terms of a meta-game being played by the designers of the individual Q-learners who set their agents' parameters. Indeed, Nash equilibria of the meta-game are characterized by heterogeneous parameters, and resulting outcomes achieve little to no cooperation beyond Nash. In conclusion, we suggest a novel perspective for thinking about regulation and collusion, and discuss the implications of our results for Bertrand oligopoly pricing games.

arXiv:2502.17666

Yes, Q-learning Helps Offline In-Context RL

Existing offline in-context reinforcement learning (ICRL) methods have predominantly relied on supervised training objectives, which are known to have limitations in offline RL settings. In this study, we explore the integration of RL objectives within an offline ICRL framework. Through experiments on more than 150 GridWorld and MuJoCo environment-derived datasets, we demonstrate that optimizing RL objectives directly improves performance by approximately 30% on average compared to widely adopted Algorithm Distillation (AD), across various dataset coverages, structures, expertise levels, and environmental complexities. Furthermore, in the challenging XLand-MiniGrid environment, RL objectives doubled the performance of AD. Our results also reveal that the addition of conservatism during value learning brings additional improvements in almost all settings tested. Our findings emphasize the importance of aligning ICRL learning objectives with the RL reward-maximization goal, and demonstrate that offline RL is a promising direction for advancing ICRL.

arXiv:2502.17538

Policy Learning with a Natural Language Action Space: A Causal Approach

This paper introduces a novel causal framework for multi-stage decision-making in natural language action spaces where outcomes are only observed after a sequence of actions. While recent approaches like Proximal Policy Optimization (PPO) can handle such delayed-reward settings in high-dimensional action spaces, they typically require multiple models (policy, value, and reward) and substantial training data. Our approach employs Q-learning to estimate Dynamic Treatment Regimes (DTR) through a single model, enabling data-efficient policy learning via gradient ascent on language embeddings. A key technical contribution of our approach is a decoding strategy that translates optimized embeddings back into coherent natural language. We evaluate our approach on mental health intervention, hate speech countering, and sentiment transfer tasks, demonstrating significant improvements over competitive baselines across multiple metrics. Notably, our method achieves superior transfer strength while maintaining content preservation and fluency, as validated through human evaluation. Our work provides a practical foundation for learning optimal policies in complex language tasks where training data is limited.

arXiv:2502.15792

Multi-Objective Reinforcement Learning for Critical Scenario Generation of Autonomous Vehicles

Autonomous vehicles (AVs) make driving decisions without human intervention. Therefore, ensuring AVs' dependability is critical. Despite significant research and development in AV development, their dependability assurance remains a significant challenge due to the complexity and unpredictability of their operating environments. Scenario-based testing evaluates AVs under various driving scenarios, but the unlimited number of potential scenarios highlights the importance of identifying critical scenarios that can violate safety or functional requirements. Such requirements are inherently interdependent and need to be tested simultaneously. To this end, we propose MOEQT, a novel multi-objective reinforcement learning (MORL)-based approach to generate critical scenarios that simultaneously test interdependent safety and functional requirements. MOEQT adapts Envelope Q-learning as the MORL algorithm, which dynamically adapts multi-objective weights to balance the relative importance between multiple objectives. MOEQT generates critical scenarios to violate multiple requirements through dynamically interacting with the AV environment, ensuring comprehensive AV testing. We evaluate MOEQT using an advanced end-to-end AV controller and a high-fidelity simulator and compare MOEQT with two baselines: a random strategy and a single-objective RL with a weighted reward function. Our evaluation results show that MOEQT achieved an overall better performance in identifying critical scenarios for violating multiple requirements than the baselines.