arXiv subjects

Huizhen Yu

Publications and source records attributed to Huizhen Yu.

At least 19 recordsLinked to original sources

arXiv2026

Quasi-multimodal-based pathophysiological feature learning for retinal disease diagnosis

Retinal diseases spanning a broad spectrum can be effectively identified and diagnosed using complementary signals from multimodal data. However, multimodal diagnosis in ophthalmic practice is typically challenged in terms of data heterogeneity, potential invasiveness, registration complexity, and so on. As such, a unified framework that integrates multimodal data synthesis and fusion is proposed for retinal disease classification and grading. Specifically, the synthesized multimodal data incorporates fundus fluorescein angiography (FFA), multispectral imaging (MSI), and saliency maps that emphasize latent lesions as well as optic disc/cup regions. Parallel models are independently trained to learn modality-specific representations that capture cross-pathophysiological signatures. These features are then adaptively calibrated within and across modalities to perform information pruning and flexible integration according to downstream tasks. The proposed learning system is thoroughly interpreted through visualizations in both image and feature spaces. Extensive experiments on two public datasets demonstrated the superiority of our approach over state-of-the-art ones in the tasks of multi-label classification (F1-score: 0.683, AUC: 0.953) and diabetic retinopathy grading (Accuracy:0.842, Kappa: 0.861). This work not only enhances the accuracy and efficiency of retinal disease screening but also offers a scalable framework for data augmentation across various medical imaging modalities.

cs.CV↗

arXiv2025

Average-reward reinforcement learning in semi-Markov decision processes via relative value iteration

This paper applies the authors' recent results on asynchronous stochastic approximation (SA) in the Borkar-Meyn framework to reinforcement learning in average-reward semi-Markov decision processes (SMDPs). We establish the convergence of an asynchronous SA analogue of Schweitzer's classical relative value iteration algorithm, RVI Q-learning, for finite-space, weakly communicating SMDPs. In particular, we show that the algorithm converges almost surely to a compact, connected subset of solutions to the average-reward optimality equation, with convergence to a unique, sample path-dependent solution under additional stepsize and asynchrony conditions. Moreover, to make full use of the SA framework, we introduce new monotonicity conditions for estimating the optimal reward rate in RVI Q-learning. These conditions substantially expand the previously considered algorithmic framework and are addressed through novel arguments in the stability and convergence analysis of RVI Q-learning.

cs.LG↗

arXiv2024

Asynchronous Stochastic Approximation with Applications to Average-Reward Reinforcement Learning

This paper investigates the stability and convergence properties of asynchronous stochastic approximation (SA) algorithms, with a focus on extensions relevant to average-reward reinforcement learning. We first extend a stability proof method of Borkar and Meyn to accommodate more general noise conditions than previously considered, thereby yielding broader convergence guarantees for asynchronous SA. To sharpen the convergence analysis, we further examine the shadowing properties of asynchronous SA, building on a dynamical systems approach of Hirsch and Bena\"{i}m. These results provide a theoretical foundation for a class of relative value iteration-based reinforcement learning algorithms -- developed and analyzed in a companion paper -- for solving average-reward Markov and semi-Markov decision processes.

cs.LG↗

arXiv2024

On Convergence of Average-Reward Q-Learning in Weakly Communicating Markov Decision Processes

This paper analyzes reinforcement learning (RL) algorithms for Markov decision processes (MDPs) under the average-reward criterion. We focus on Q-learning algorithms based on relative value iteration (RVI), which are model-free stochastic analogues of the classical RVI method for average-reward MDPs. These algorithms have low per-iteration complexity, making them well-suited for large state space problems. We extend the almost-sure convergence analysis of RVI Q-learning algorithms developed by Abounadi, Bertsekas, and Borkar (2001) from unichain to weakly communicating MDPs. This extension is important both practically and theoretically: weakly communicating MDPs cover a much broader range of applications compared to unichain MDPs, and their optimality equations have a richer solution structure (with multiple degrees of freedom), introducing additional complexity in proving algorithmic convergence. We also characterize the sets to which RVI Q-learning algorithms converge, showing that they are compact, connected, potentially nonconvex, and comprised of solutions to the average-reward optimality equation, with exactly one less degree of freedom than the general solution set of this equation. Furthermore, we extend our analysis to two RVI-based hierarchical average-reward RL algorithms using the options framework, proving their almost-sure convergence and characterizing their sets of convergence under the assumption that the underlying semi-Markov decision process is weakly communicating.

cs.LG↗

arXiv2023

A Note on Stability in Asynchronous Stochastic Approximation without Communication Delays

In this paper, we study asynchronous stochastic approximation algorithms without communication delays. Our main contribution is a stability proof for these algorithms that extends a method of Borkar and Meyn by accommodating more general noise conditions. We also derive convergence results from this stability result and discuss their application in important average-reward reinforcement learning problems.

cs.LG↗

arXiv2022

On Strategic Measures and Optimality Properties in Discrete-Time Stochastic Control with Universally Measurable Policies

This paper concerns discrete-time infinite-horizon stochastic control systems with Borel state and action spaces and universally measurable policies. We study optimization problems on strategic measures induced by the policies in these systems. The results are then applied to risk-neutral and risk-sensitive Markov decision processes, as well as their partially observable counterparts, to establish the measurability of the optimal value functions and the existence of universally measurable, randomized or nonrandomized, $\epsilon$-optimal policies, for a variety of average cost criteria and risk criteria. We also extend our analysis to a class of minimax control problems and establish similar optimality results under the axiom of analytic determinacy.

math.OC↗

arXiv2021

Average-Cost Optimality Results for Borel-Space Markov Decision Processes with Universally Measurable Policies

We consider discrete-time Markov Decision Processes with Borel state and action spaces and universally measurable policies. For several long-run average cost criteria, we establish the following optimality results: the optimal average cost functions are lower semianalytic, there exist universally measurable semi-Markov or history-dependent $\epsilon$-optimal policies, and similar results hold for the minimum average costs achievable by Markov or stationary policies. We then analyze the structure of the optimal average cost functions, proving sufficient conditions for them to be constant almost everywhere with respect to certain $\sigma$-finite measures. The most important condition here is that each subset of states with positive measure be reachable with probability one under some policy. We obtain our results by exploiting an inequality for the optimal average cost functions and its connection with submartingales, and, in a special case that involves stationary policies, also by using the theory of recurrent Markov chains.

math.OC↗

arXiv2020

Average Cost Optimality Inequality for Markov Decision Processes with Borel Spaces and Universally Measurable Policies

We consider average-cost Markov decision processes (MDPs) with Borel state and action spaces and universally measurable policies. For the nonnegative cost model and an unbounded cost model with a Lyapunov-type stability character, we introduce a set of new conditions under which we prove the average cost optimality inequality (ACOI) via the vanishing discount factor approach. Unlike most existing results on the ACOI, our result does not require any compactness and continuity conditions on the MDPs. Instead, the main idea is to use the almost-uniform-convergence property of a pointwise convergent sequence of measurable functions as asserted in Egoroff's theorem. Our conditions are formulated in order to exploit this property. Among others, we require that for each state, on selected subsets of actions at that state, the state transition stochastic kernel is majorized by finite measures. We combine this majorization property of the transition kernel with Egoroff's theorem to prove the ACOI.

math.OC↗

arXiv2019

On Linear Programming for Constrained and Unconstrained Average-Cost Markov Decision Processes with Countable Action Spaces and Strictly Unbounded Costs

We consider the linear programming approach for constrained and unconstrained Markov decision processes (MDPs) under the long-run average cost criterion, where the class of MDPs in our study have Borel state spaces and discrete countable action spaces. Under a strict unboundedness condition on the one-stage costs and a recently introduced majorization condition on the state transition stochastic kernel, we study infinite-dimensional linear programs for the average-cost MDPs and prove the absence of a duality gap and other optimality results. Our results do not require a lower-semicontinuous MDP model. Thus, they can be applied to countable action space MDPs where the dynamics and one-stage costs are discontinuous in the state variable. Our proofs make use of the continuity property of Borel measurable functions asserted by Lusin's theorem.

math.OC↗

arXiv2019

On the Minimum Pair Approach for Average-Cost Markov Decision Processes with Countable Discrete Action Spaces and Strictly Unbounded Costs

We consider average-cost Markov decision processes (MDPs) with Borel state spaces, countable, discrete action spaces, and strictly unbounded one-stage costs. For the minimum pair approach, we introduce a new majorization condition on the state transition stochastic kernel, in place of the commonly required continuity conditions on the MDP model. We combine this majorization condition with Lusin's theorem to prove the existence of a stationary minimum pair, i.e., a stationary policy paired with an invariant probability measure induced on the state space, with the property that the pair attains the minimum long-run average cost over all policies and initial distributions. We also establish other optimality properties of a stationary minimum pair, and for the stationary policy in such a pair, under additional recurrence or regularity conditions, we prove its pathwise optimality and strong optimality. Our results can be applied to a class of countable action space MDPs in which the dynamics and one-stage costs are discontinuous with respect to the state variable.

math.OC↗

arXiv2019

On Markov Decision Processes with Borel Spaces and an Average Cost Criterion

We consider average-cost Markov decision processes (MDPs) with Borel state and action spaces and universally measurable policies. For the nonnegative cost model and an unbounded cost model, we introduce a set of conditions under which we prove the average cost optimality inequality (ACOI) via the vanishing discount factor approach. Unlike most existing results on the ACOI, which require compactness/continuity conditions on the MDP, our result does not and can be applied to problems with discontinuous dynamics and one-stage costs. The key idea here is to replace the compactness/continuity conditions used in the prior work by what we call majorization type conditions. In particular, among others, we require that for each state, on selected subsets of actions at that state, the state transition stochastic kernel is majorized by finite measures, and we use this majorization property together with Egoroff's theorem to prove the ACOI. We also consider the minimum pair approach for average-cost MDPs and apply the majorization idea. For the case of a discrete action space and strictly unbounded costs, we prove the existence of a minimum pair that consists of a stationary policy and an invariant probability measure induced by the policy. This result is derived by combining Lusin's theorem with another majorization condition we introduce, and it can be applied to a class of countable action space MDPs in which, with respect to the state variable, the dynamics and one-stage costs are discontinuous.

math.OC↗

arXiv2018

Two geometric input transformation methods for fast online reinforcement learning with neural nets

We apply neural nets with ReLU gates in online reinforcement learning. Our goal is to train these networks in an incremental manner, without the computationally expensive experience replay. By studying how individual neural nodes behave in online training, we recognize that the global nature of ReLU gates can cause undesirable learning interference in each node's learning behavior. We propose reducing such interferences with two efficient input transformation methods that are geometric in nature and match well the geometric property of ReLU gates. The first one is tile coding, a classic binary encoding scheme originally designed for local generalization based on the topological structure of the input space. The second one (EmECS) is a new method we introduce; it is based on geometric properties of convex sets and topological embedding of the input space into the boundary of a convex set. We discuss the behavior of the network when it operates on the transformed inputs. We also compare it experimentally with some neural nets that do not use the same input transformations, and with the classic algorithm of tile coding plus a linear function approximator, and on several online reinforcement learning tasks, we show that the neural net with tile coding or EmECS can achieve not only faster learning but also more accurate approximations. Our results strongly suggest that geometric input transformation of this type can be effective for interference reduction and takes us a step closer to fully incremental reinforcement learning with neural nets.

cs.LG↗

arXiv2017

On Convergence of some Gradient-based Temporal-Differences Algorithms for Off-Policy Learning

We consider off-policy temporal-difference (TD) learning methods for policy evaluation in Markov decision processes with finite spaces and discounted reward criteria, and we present a collection of convergence results for several gradient-based TD algorithms with linear function approximation. The algorithms we analyze include: (i) two basic forms of two-time-scale gradient-based TD algorithms, which we call GTD and which minimize the mean squared projected Bellman error using stochastic gradient-descent; (ii) their "robustified" biased variants; (iii) their mirror-descent versions which combine the mirror-descent idea with TD learning; and (iv) a single-time-scale version of GTD that solves minimax problems formulated for approximate policy evaluation. We derive convergence results for three types of stepsizes: constant stepsize, slowly diminishing stepsize, as well as the standard type of diminishing stepsize with a square-summable condition. For the first two types of stepsizes, we apply the weak convergence method from stochastic approximation theory to characterize the asymptotic behavior of the algorithms, and for the standard type of stepsize, we analyze the algorithmic behavior with respect to a stronger mode of convergence, almost sure convergence. Our convergence results are for the aforementioned TD algorithms with three general ways of setting their $\lambda$-parameters: (i) state-dependent $\lambda$; (ii) a recently proposed scheme of using history-dependent $\lambda$ to keep the eligibility traces of the algorithms bounded while allowing for relatively large values of $\lambda$; and (iii) a composite scheme of setting the $\lambda$-parameters that combines the preceding two schemes and allows a broader class of generalized Bellman operators to be used for approximate policy evaluation with TD methods.

cs.LG↗

arXiv2017

On Generalized Bellman Equations and Temporal-Difference Learning

We consider off-policy temporal-difference (TD) learning in discounted Markov decision processes, where the goal is to evaluate a policy in a model-free way by using observations of a state process generated without executing the policy. To curb the high variance issue in off-policy TD learning, we propose a new scheme of setting the $\lambda$-parameters of TD, based on generalized Bellman equations. Our scheme is to set $\lambda$ according to the eligibility trace iterates calculated in TD, thereby easily keeping these traces in a desired bounded range. Compared with prior work, this scheme is more direct and flexible, and allows much larger $\lambda$ values for off-policy TD learning with bounded traces. As to its soundness, using Markov chain theory, we prove the ergodicity of the joint state-trace process under nonrestrictive conditions, and we show that associated with our scheme is a generalized Bellman equation (for the policy to be evaluated) that depends on both the evolution of $\lambda$ and the unique invariant probability measure of the state-trace process. These results not only lead immediately to a characterization of the convergence behavior of least-squares based implementation of our scheme, but also prepare the ground for further analysis of gradient-based implementations.

cs.LG↗

arXiv2017

Multi-step Off-policy Learning Without Importance Sampling Ratios

To estimate the value functions of policies from exploratory data, most model-free off-policy algorithms rely on importance sampling, where the use of importance sampling ratios often leads to estimates with severe variance. It is thus desirable to learn off-policy without using the ratios. However, such an algorithm does not exist for multi-step learning with function approximation. In this paper, we introduce the first such algorithm based on temporal-difference (TD) learning updates. We show that an explicit use of importance sampling ratios can be eliminated by varying the amount of bootstrapping in TD updates in an action-dependent manner. Our new algorithm achieves stability using a two-timescale gradient-based TD update. A prior algorithm based on lookup table representation called Tree Backup can also be retrieved using action-dependent bootstrapping, becoming a special case of our algorithm. In two challenging off-policy tasks, we demonstrate that our algorithm is stable, effectively avoids the large variance issue, and can perform substantially better than its state-of-the-art counterpart.

cs.LG↗

arXiv2016

Some Simulation Results for Emphatic Temporal-Difference Learning Algorithms

This is a companion note to our recent study of the weak convergence properties of constrained emphatic temporal-difference learning (ETD) algorithms from a theoretic perspective. It supplements the latter analysis with simulation results and illustrates the behavior of some of the ETD algorithms using three example problems.

cs.LG↗

arXiv2015

Weak Convergence Properties of Constrained Emphatic Temporal-difference Learning with Constant and Slowly Diminishing Stepsize

We consider the emphatic temporal-difference (TD) algorithm, ETD($\lambda$), for learning the value functions of stationary policies in a discounted, finite state and action Markov decision process. The ETD($\lambda$) algorithm was recently proposed by Sutton, Mahmood, and White to solve a long-standing divergence problem of the standard TD algorithm when it is applied to off-policy training, where data from an exploratory policy are used to evaluate other policies of interest. The almost sure convergence of ETD($\lambda$) has been proved in our recent work under general off-policy training conditions, but for a narrow range of diminishing stepsize. In this paper we present convergence results for constrained versions of ETD($\lambda$) with constant stepsize and with diminishing stepsize from a broad range. Our results characterize the asymptotic behavior of the trajectory of iterates produced by those algorithms, and are derived by combining key properties of ETD($\lambda$) with powerful convergence theorems from the weak convergence methods in stochastic approximation theory. For the case of constant stepsize, in addition to analyzing the behavior of the algorithms in the limit as the stepsize parameter approaches zero, we also analyze their behavior for a fixed stepsize and bound the deviations of their averaged iterates from the desired solution. These results are obtained by exploiting the weak Feller property of the Markov chains associated with the algorithms, and by using ergodic theorems for weak Feller Markov chains, in conjunction with the convergence results we get from the weak convergence methods. Besides ETD($\lambda$), our analysis also applies to the off-policy TD($\lambda$) algorithm, when the divergence issue is avoided by setting $\lambda$ sufficiently large.

cs.LG↗

arXiv2015

Emphatic Temporal-Difference Learning

Emphatic algorithms are temporal-difference learning algorithms that change their effective state distribution by selectively emphasizing and de-emphasizing their updates on different time steps. Recent works by Sutton, Mahmood and White (2015), and Yu (2015) show that by varying the emphasis in a particular way, these algorithms become stable and convergent under off-policy training with linear function approximation. This paper serves as a unified summary of the available results from both works. In addition, we demonstrate the empirical benefits from the flexibility of emphatic algorithms, including state-dependent discounting, state-dependent bootstrapping, and the user-specified allocation of function approximation resources.

cs.LG↗