Markov decision processes: a tool for sequential decision making under uncertainty
PulseAugur coverage of Markov decision processes: a tool for sequential decision making under uncertainty — every cluster mentioning Markov decision processes: a tool for sequential decision making under uncertainty across labs, papers, and developer communities, ranked by signal.
12 day(s) with sentiment data
-
New research details $O(1/n)$ convergence rate for RVI Q-learning
Researchers have established a finite-time convergence guarantee for synchronous relative value iteration (RVI) Q-learning in finite weakly communicating Markov decision processes (MDPs). The study details convergence r…
-
New Q-learning algorithm tackles model uncertainty in decision processes
Researchers have developed a new Q-learning algorithm designed for Markov decision processes that specifically addresses model uncertainty. This algorithm operates within a finite ambiguity set, allowing for user-define…
-
New PRISM framework improves AI in online medical consultations
Researchers have developed PRISM, a novel framework for online medical consultations that addresses the challenge of irregular patient disclosures and topic shifts. Inspired by partially observable Markov decision proce…
-
New algorithm learns sequential decisions from diverse data sources
Researchers have developed a new algorithm called Group-Robust Pessimistic Value Iteration (GR-PEVI) to learn sequential decision-making policies from multiple, potentially heterogeneous data sources. This method addres…
-
arXiv sees new research on lifted reasoning and robust RL
Two new research papers were submitted to arXiv on October 26, 2026, focusing on distinct areas of AI. The first paper, "Query-Driven Learning for Lifted Reasoning," explores integrating learning and reasoning in first-…
-
New InCOPE method improves off-policy evaluation with limited data
Researchers have developed InCOPE, a novel method for In-Context Off-Policy Evaluation (OPE) designed to accurately assess decision policies using limited behavior-policy data. The approach involves pretraining a networ…
-
LLM Agents' Memory Systems Face New Benchmarks for Safety and Efficiency
Recent research explores vulnerabilities and improvements in Large Language Model (LLM) agent memory systems. Studies highlight how memory formation can be susceptible to poisoning, leading to unreliable agent behavior …
-
New adaptive RMDP framework enhances AI surveillance missions
Researchers have developed an adaptive framework for Robust Markov Decision Processes (RMDPs) specifically designed for Collaborative Combat Aircraft (CCAs) engaged in Intelligence, Surveillance, and Reconnaissance (ISR…
-
New algorithms offer theoretical guarantees for low-rank continuous-state CMDPs
Researchers have developed two novel algorithms, REP-PD-TEG and REP-PD-H-TEG, for representation learning in low-rank constrained Markov Decision Processes (CMDPs) with continuous state spaces. REP-PD-TEG addresses the …
-
Survey maps reinforcement learning implementations on FPGAs and ICs
This survey paper reviews over 100 studies on the implementation of reinforcement learning (RL) algorithms on field-programmable gate arrays (FPGAs) and integrated circuits (ICs). It categorizes research by RL algorithm…
-
New research explores Bellman optimality for average-reward robust Markov Decision Processes
This paper delves into the theoretical underpinnings of average-reward robust Markov Decision Processes (MDPs), a complex area of sequential decision-making. It focuses on the constant-gain setting to establish a genera…
-
Survey maps reinforcement learning implementations on FPGAs and ICs
This survey examines over 100 studies on the implementation of reinforcement learning (RL) algorithms on field-programmable gate arrays (FPGAs) and integrated circuits (ICs). It categorizes research by RL family, hardwa…
-
New research explores parametric resonance in primal-dual learning for constrained MDPs
Researchers have analyzed primal-dual methods for constrained Markov decision processes, considering scenarios with slow non-stationarity. The study reveals that in additive channels where constraint levels oscillate, t…
-
V-SHARE algorithm offers provable convergence for reinforcement learning
Researchers have introduced V-SHARE, a novel advantage-based value-learning algorithm for reinforcement learning that enhances information sharing across actions. This method decomposes the action-value function into st…
-
New Q-learning algorithms achieve fast rates with KL regularization
Researchers have developed new algorithms for offline Q-learning in finite-horizon Markov decision processes, incorporating Kullback-Leibler (KL) regularization. The proposed KL-regularized Q-Learning (KLQ) algorithm ac…
-
New method embeds information in AI policy randomness
Researchers have introduced a novel method called posterior-matching randomness modulation (PRAM) to embed information within the random actions of a stochastic policy in Markov Decision Processes (MDPs). This technique…
-
New CauCRL Method Improves Goal-Conditioned RL with Confounded Data
Researchers have developed a new method called Causal Contrastive Reinforcement Learning (CauCRL) to address challenges in goal-conditioned reinforcement learning when offline data is confounded. CauCRL introduces count…
-
New method improves reinforcement learning efficiency via in-context modeling
Researchers have developed a new method to improve the efficiency of reinforcement learning (RL) by addressing limitations in risk predictive models (RPMs). Their approach recasts risk prediction as an in-context sequen…
-
New framework tackles uncertainty in AI decision-making
Researchers have developed a new framework for counterfactual inference in Markov Decision Processes (MDPs) that accounts for both statistical and causal uncertainties. This approach provides analytical bounds for count…
-
New research tackles optimal regret in adversarial MDPs · 2 papers
Two new research papers submitted to arXiv explore advancements in adversarial Markov decision processes (MDPs). The first paper introduces MA-OPS, an algorithm designed to achieve optimal regret while adhering to stoch…