Temporal difference learning (TD) is a simple iterative algorithm used to estimate the value function corresponding to a given policy in a Markov decision process. Although TD is one of the most widely used algorithms in reinforcement learning, its theoretical analysis has proved challenging and few guarantees on its statistical efficiency are available. In this work, we provide a simple and explicit finite time analysis of temporal difference learning with linear function approximation. Except for a few key insights, our analysis mirrors standard techniques for analyzing stochastic gradient descent algorithms, and therefore inherits the simplicity and elegance of that literature. Final sections of the paper show how all of our main results extend to the study of TD learning with eligibility traces, known as TD($λ$), and to Q-learning applied in high-dimensional optimal stopping problems.
Temporal difference learning with linear function approximation is a popular method to obtain a low-dimensional approximation of the value function of a policy in a Markov Decision Process. We give a new interpretation of this method in terms of a splitting of the gradient of an appropriately chosen function. As a consequence of this interpretation, convergence proofs for gradient descent can be applied almost verbatim to temporal difference learning. Beyond giving a new, fuller explanation of why temporal difference works, our interpretation also yields improved convergence times. We consider…
To act and plan in complex environments, we posit that agents should have a mental simulator of the world with three characteristics: (a) it should build an abstract state representing the condition of the world; (b) it should form a belief which represents uncertainty on the world; (c) it should go beyond simple step-by-step simulation, and exhibit temporal abstraction. Motivated by the absence of a model satisfying all these requirements, we propose TD-VAE, a generative sequence model that learns representations containing explicit beliefs about states several steps into the future, and that…
A common optimization tool used in deep reinforcement learning is momentum, which consists in accumulating and discounting past gradients, reapplying them at each iteration. We argue that, unlike in supervised learning, momentum in Temporal Difference (TD) learning accumulates gradients that become doubly stale: not only does the gradient of the loss change due to parameter updates, the loss itself changes due to bootstrapping. We first show that this phenomenon exists, and then propose a first-order correction term to momentum. We show that this correction term improves sample efficiency in p…
We consider the off-policy evaluation problem in Markov decision processes with function approximation. We propose a generalization of the recently introduced \emph{emphatic temporal differences} (ETD) algorithm \citep{SuttonMW15}, which encompasses the original ETD($λ$), as well as several other off-policy evaluation algorithms as special cases. We call this framework \ETD, where our introduced parameter $β$ controls the decay rate of an importance-sampling term. We study conditions under which the projected fixed-point equation underlying \ETD\ involves a contraction operator, allowing us to…
Learning the value function of a given policy (target policy) from the data samples obtained from a different policy (behavior policy) is an important problem in Reinforcement Learning (RL). This problem is studied under the setting of off-policy prediction. Temporal Difference (TD) learning algorithms are a popular class of algorithms for solving the prediction problem. TD algorithms with linear function approximation are shown to be convergent when the samples are generated from the target policy (known as on-policy prediction). However, it has been well established in the literature that of…
Reinforcement learning (RL) has been successfully used to solve many continuous control tasks. Despite its impressive results however, fundamental questions regarding the sample complexity of RL on continuous problems remain open. We study the performance of RL in this setting by considering the behavior of the Least-Squares Temporal Difference (LSTD) estimator on the classic Linear Quadratic Regulator (LQR) problem from optimal control. We give the first finite-time analysis of the number of samples needed to estimate the value function for a fixed static state-feedback policy to within $\var…
Off-policy reinforcement learning has many applications including: learning from demonstration, learning multiple goal seeking policies in parallel, and representing predictive knowledge. Recently there has been an proliferation of new policy-evaluation algorithms that fill a longstanding algorithmic void in reinforcement learning: combining robustness to off-policy sampling, function approximation, linear complexity, and temporal difference (TD) updates. This paper contains two main contributions. First, we derive two new hybrid TD policy-evaluation algorithms, which fill a gap in this collec…
We consider reinforcement learning with performance evaluated by a dynamic risk measure. We construct a projected risk-averse dynamic programming equation and study its properties. Then we propose risk-averse counterparts of the methods of temporal differences and we prove their convergence with probability one. We also perform an empirical study on a complex transportation problem.
Q-learning is a reliable but inefficient off-policy temporal-difference method, backing up reward only one step at a time. Replacing traces, using a recency heuristic, are more efficient but less reliable. In this work, we introduce model-free, off-policy temporal difference methods that make better use of experience than Watkins' Q(λ). We introduce both Optimistic Q(λ) and the temporal second difference trace (TSDT). TSDT is particularly powerful in deterministic domains. TSDT uses neither recency nor frequency heuristics, storing (s,a,r,s',δ) so that off-policy updates can be performed after…
Multi-step temporal difference (TD) learning is an important approach in reinforcement learning, as it unifies one-step TD learning with Monte Carlo methods in a way where intermediate algorithms can outperform either extreme. They address a bias-variance trade off between reliance on current estimates, which could be poor, and incorporating longer sampled reward sequences into the updates. Especially in the off-policy setting, where the agent aims to learn about a policy different from the one generating its behaviour, the variance in the updates can cause learning to diverge as the number of…
We address the problem of policy evaluation in discounted Markov decision processes, and provide instance-dependent guarantees on the $\ell_\infty$-error under a generative model. We establish both asymptotic and non-asymptotic versions of local minimax lower bounds for policy evaluation, thereby providing an instance-dependent baseline by which to compare algorithms. Theory-inspired simulations show that the widely-used temporal difference (TD) algorithm is strictly suboptimal when evaluated in a non-asymptotic setting, even when combined with Polyak-Ruppert iterate averaging. We remedy this …
Recently, a new multi-step temporal learning algorithm, called $Q(σ)$, unifies $n$-step Tree-Backup (when $σ=0$) and $n$-step Sarsa (when $σ=1$) by introducing a sampling parameter $σ$. However, similar to other multi-step temporal-difference learning algorithms, $Q(σ)$ needs much memory consumption and computation time. Eligibility trace is an important mechanism to transform the off-line updates into efficient on-line ones which consume less memory and computation time. In this paper, we further develop the original $Q(σ)$, combine it with eligibility traces and propose a new algorithm, call…
Because reinforcement learning suffers from a lack of scalability, online value (and Q-) function approximation has received increasing interest this last decade. This contribution introduces a novel approximation scheme, namely the Kalman Temporal Differences (KTD) framework, that exhibits the following features: sample-efficiency, non-linear approximation, non-stationarity handling and uncertainty management. A first KTD-based algorithm is provided for deterministic Markov Decision Processes (MDP) which produces biased estimates in the case of stochastic transitions. Than the eXtended KTD fr…
We consider emphatic temporal-difference learning algorithms for policy evaluation in discounted Markov decision processes with finite spaces. Such algorithms were recently proposed by Sutton, Mahmood, and White (2015) as an improved solution to the problem of divergence of off-policy temporal-difference learning with linear function approximation. We present in this paper the first convergence proofs for two emphatic algorithms, ETD($λ$) and ELSTD($λ$). We prove, under general off-policy conditions, the convergence in $L^1$ for ELSTD($λ$) iterates, and the almost sure convergence of the appro…
We derive an equation for temporal difference learning from statistical principles. Specifically, we start with the variational principle and then bootstrap to produce an updating rule for discounted state value estimates. The resulting equation is similar to the standard equation for temporal difference learning with eligibility traces, so called TD(lambda), however it lacks the parameter alpha that specifies the learning rate. In the place of this free parameter there is now an equation for the learning rate that is specific to each state transition. We experimentally test this new learning …
Temporal Difference learning or TD($λ$) is a fundamental algorithm in the field of reinforcement learning. However, setting TD's $λ$ parameter, which controls the timescale of TD updates, is generally left up to the practitioner. We formalize the $λ$ selection problem as a bias-variance trade-off where the solution is the value of $λ$ that leads to the smallest Mean Squared Value Error (MSVE). To solve this trade-off we suggest applying Leave-One-Trajectory-Out Cross-Validation (LOTO-CV) to search the space of $λ$ values. Unfortunately, this approach is too computationally expensive for most p…
One of the main obstacles to broad application of reinforcement learning methods is the parameter sensitivity of our core learning algorithms. In many large-scale applications, online computation and function approximation represent key strategies in scaling up reinforcement learning algorithms. In this setting, we have effective and reasonably well understood algorithms for adapting the learning-rate parameter, online during learning. Such meta-learning approaches can improve robustness of learning and enable specialization to current task, improving learning speed. For temporal-difference le…
University of Liverpool（利物浦大学）；Centrum Wiskunde & Informatica（数学与计算机科学中心）
We present a new Q-function operator for temporal difference (TD) learning methods that explicitly encodes robustness against significant rare events (SRE) in critical domains. The operator, which we call the $κ$-operator, allows to learn a robust policy in a model-based fashion without actually observing the SRE. We introduce single- and multi-agent robust TD methods using the operator $κ$. We prove convergence of the operator to the optimal robust Q-function with respect to the model using the theory of Generalized Markov Decision Processes. In addition we prove convergence to the optimal Q-…
An effective approach to exploration in reinforcement learning is to rely on an agent's uncertainty over the optimal policy, which can yield near-optimal exploration strategies in tabular settings. However, in non-tabular settings that involve function approximators, obtaining accurate uncertainty estimates is almost as challenging a problem. In this paper, we highlight that value estimates are easily biased and temporally inconsistent. In light of this, we propose a novel method for estimating uncertainty over the value function that relies on inducing a distribution over temporal difference …
Reinforcement learning lies at the intersection of several challenges. Many applications of interest involve extremely large state spaces, requiring function approximation to enable tractable computation. In addition, the learner has only a single stream of experience with which to evaluate a large number of possible courses of action, necessitating algorithms which can learn off-policy. However, the combination of off-policy learning with function approximation leads to divergence of temporal difference methods. Recent work into gradient-based temporal difference methods has promised a path t…
Temporal-difference learning (TD), coupled with neural networks, is among the most fundamental building blocks of deep reinforcement learning. However, due to the nonlinearity in value function approximation, such a coupling leads to nonconvexity and even divergence in optimization. As a result, the global convergence of neural TD remains unclear. In this paper, we prove for the first time that neural TD converges at a sublinear rate to the global optimum of the mean-squared projected Bellman error for policy evaluation. In particular, we show how such global convergence is enabled by the over…
Faculty of Electrical Engineering, Technion（以色列理工学院电气工程学院）；H. Milton Stewart School of Industrial and Systems Engineering, Georgia Institute of Technology（佐治亚理工学院H. Milton Stewart工业与系统工程学院）
We devise a distributional variant of gradient temporal-difference (TD) learning. Distributional reinforcement learning has been demonstrated to outperform the regular one in the recent study \citep{bellemare2017distributional}. In the policy evaluation setting, we design two new algorithms called distributional GTD2 and distributional TDC using the Cram{é}r distance on the distributional version of the Bellman error objective function, which inherits advantages of both the nonlinear gradient TD algorithms and the distributional RL approach. In the control setting, we propose the distributiona…
Improved High-Probability Bounds for the Temporal Difference Learning Algorithm via Exponential Stability
HSE University（高等经济大学）；Centre de Mathématiques Appliquées（应用数学中心）；CNRS（法国国家科学研究中心）；École polytechnique（巴黎综合理工学院）；Institut Polytechnique de Paris（巴黎理工学院）；Université Paris-Saclay（巴黎萨克雷大学）；Laboratoire de mathématiques d’Orsay（奥尔赛数学实验室）；Steklov Mathematical Institute of Russian Academy of Sciences（俄罗斯科学院斯捷克洛夫数学研究所）；Mohamed bin Zayed University of Artificial Intelligence（穆罕默德·本·扎耶德人工智能大学）
We consider the core reinforcement-learning problem of on-policy value function approximation from a batch of trajectory data, and focus on various issues of Temporal Difference (TD) learning and Monte Carlo (MC) policy evaluation. The two methods are known to achieve complementary bias-variance trade-off properties, with TD tending to achieve lower variance but potentially higher bias. In this paper, we argue that the larger bias of TD can be a result of the amplification of local approximation errors. We address this by proposing an algorithm that adaptively switches between TD and MC in eac…
Emphatic Temporal Difference (TD) methods are a class of off-policy Reinforcement Learning (RL) methods involving the use of followon traces. Despite the theoretical success of emphatic TD methods in addressing the notorious deadly triad of off-policy RL, there are still two open problems. First, followon traces typically suffer from large variance, making them hard to use in practice. Second, though Yu (2015) confirms the asymptotic convergence of some emphatic TD methods for prediction problems, there is still no finite sample analysis for any emphatic TD method for prediction, much less con…
Gradient temporal difference (Gradient TD) algorithms are a popular class of stochastic approximation (SA) algorithms used for policy evaluation in reinforcement learning. Here, we consider Gradient TD algorithms with an additional heavy ball momentum term and provide choice of step size and momentum parameter that ensures almost sure convergence of these algorithms asymptotically. In doing so, we decompose the heavy ball Gradient TD iterates into three separate iterates with different step sizes. We first analyze these iterates under one-timescale SA setting using results from current literat…
Using insight from numerical approximation of ODEs and the problem formulation and solution methodology of TD learning through a Galerkin relaxation, I propose a new class of TD learning algorithms. After applying the improved numerical methods, the parameter being approximated has a guaranteed order of magnitude reduction in the Taylor Series error of the solution to the ODE for the parameter $θ(t)$ that is used in constructing the linearly parameterized value function. Predictor-Corrector Temporal Difference (PCTD) is what I call the translated discrete time Reinforcement Learning(RL) algori…
Understanding the affective, cognitive and behavioural processes involved in risk taking is essential for treatment and for setting environmental conditions to limit damage. Using Temporal Difference Reinforcement Learning (TDRL) we computationally investigated the effect of optimism in risk perception in a variety of goal-oriented tasks. Optimism in risk perception was studied by varying the calculation of the Temporal Difference error, i.e., delta, in three ways: realistic (stochastically correct), optimistic (assuming action control), and overly optimistic (assuming outcome control). We sho…
Northeastern University（东北大学）；Khoury College of Computer Sciences（库利计算机科学学院）
Q($σ$) is a recently proposed temporal-difference learning method that interpolates between learning from expected backups and sampled backups. It has been shown that intermediate values for the interpolation parameter $σ\in [0,1]$ perform better in practice, and therefore it is commonly believed that $σ$ functions as a bias-variance trade-off parameter to achieve these improvements. In our work, we disprove this notion, showing that the choice of $σ=0$ minimizes variance without increasing bias. This indicates that $σ$ must have some other effect on learning that is not fully understood. As a…
A central challenge to applying many off-policy reinforcement learning algorithms to real world problems is the variance introduced by importance sampling. In off-policy learning, the agent learns about a different policy than the one being executed. To account for the difference importance sampling ratios are often used, but can increase variance in the algorithms and reduce the rate of learning. Several variations of importance sampling have been proposed to reduce variance, with per-decision importance sampling being the most popular. However, the update rules for most off-policy algorithms…
University of Alberta（阿尔伯塔大学）；University of Toronto（多伦多大学）；Huawei Technologies Canada, Ltd.（加拿大华为技术有限公司）
We explore fixed-horizon temporal difference (TD) methods, reinforcement learning algorithms for a new kind of value function that predicts the sum of rewards over a $\textit{fixed}$ number of future time steps. To learn the value function for horizon $h$, these algorithms bootstrap from the value function for horizon $h-1$, or some shorter horizon. Because no value function bootstraps from itself, fixed-horizon methods are immune to the stability problems that plague other off-policy TD methods using function approximation (also known as "the deadly triad"). Although fixed-horizon methods req…
Temporal Difference Learning for High-Dimensional PIDEs with Jumps
Yau Mathematical Sciences Center, Tsinghua University（清华大学丘成桐数学科学中心）；School of Mathematics and Statistics, The University of Melbourne（墨尔本大学数学与统计学院）；Department of Mathematics, University of California, Santa Barbara（加州大学圣巴巴拉分校数学系）；Beijing Institute of Mathematical Sciences and Applications（北京数学科学与应用研究院）
Although it is well known that exploration plays a key role in Reinforcement Learning (RL), prevailing exploration strategies for continuous control tasks in RL are mainly based on naive isotropic Gaussian noise regardless of the causality relationship between action space and the task and consider all dimensions of actions equally important. In this work, we propose to conduct interventions on the primal action space to discover the causal relationship between the action space and the task reward. We propose the method of State-Wise Action Refined (SWAR), which addresses the issue of action s…
University of Illinois at Urbana-Champaign（伊利诺伊大学厄巴纳-香槟分校）
In this paper, we consider the policy evaluation problem in multi-agent reinforcement learning (MARL) and derive exact closed-form formulas for the finite-time mean-squared estimation errors of decentralized temporal difference (TD) learning with linear function approximation. Our analysis hinges upon the fact that the decentralized TD learning method can be viewed as a Markov jump linear system (MJLS). Then standard MJLS theory can be applied to quantify the mean and covariance matrix of the estimation error of the decentralized TD method at every time step. Various implications of our exact …
We propose a new approach to value function approximation which combines linear temporal difference reinforcement learning with subspace identification. In practical applications, reinforcement learning (RL) is complicated by the fact that state is either high-dimensional or partially observable. Therefore, RL methods are designed to work with features of state rather than state itself, and the success or failure of learning is often determined by the suitability of the selected features. By comparison, subspace identification (SSID) methods are designed to select a feature set which preserves…
University of Southern California（南加州大学）；Princeton University（普林斯顿大学）；Northwestern University（西北大学）
We study the policy evaluation problem in multi-agent reinforcement learning where a group of agents, with jointly observed states and private local actions and rewards, collaborate to learn the value function of a given policy via local computation and communication over a connected undirected network. This problem arises in various large-scale multi-agent systems, including power grids, intelligent transportation systems, wireless sensor networks, and multi-agent robotics. When the dimension of state-action space is large, the temporal-difference learning with linear function approximation i…
University of Alberta（阿尔伯塔大学）；Alberta Machine Intelligence Institute (Amii)（阿尔伯塔机器智能研究所（Amii））
The policy gradient theorem (Sutton et al., 2000) prescribes the usage of a cumulative discounted state distribution under the target policy to approximate the gradient. Most algorithms based on this theorem, in practice, break this assumption, introducing a distribution shift that can cause the convergence to poor solutions. In this paper, we propose a new approach of reconstructing the policy gradient from the start state without requiring a particular sampling strategy. The policy gradient calculation in this form can be simplified in terms of a gradient critic, which can be recursively est…
Sainsbury Wellcome Centre（塞恩斯伯里康惠中心）；Institute of Cognitive Neuroscience, University College London（伦敦大学学院认知神经科学研究所）；DeepMind（深度思维）
The effectiveness of Reinforcement Learning (RL) depends on an animal's ability to assign credit for rewards to the appropriate preceding stimuli. One aspect of understanding the neural underpinnings of this process involves understanding what sorts of stimulus representations support generalisation. The Successor Representation (SR), which enforces generalisation over states that predict similar outcomes, has become an increasingly popular model in this space of inquiries. Another dimension of credit assignment involves understanding how animals handle uncertainty about learned associations, …
Gradient descent or its variants are popular in training neural networks. However, in deep Q-learning with neural network approximation, a type of reinforcement learning, gradient descent (also known as Residual Gradient (RG)) is barely used to solve Bellman residual minimization problem. On the contrary, Temporal Difference (TD), an incomplete gradient descent method prevails. In this work, we perform extensive experiments to show that TD outperforms RG, that is, when the training leads to a small Bellman residual error, the solution found by TD has a better policy and is more robust against …
Temporal difference (TD) learning is an important approach in reinforcement learning, as it combines ideas from dynamic programming and Monte Carlo methods in a way that allows for online and incremental model-free learning. A key idea of TD learning is that it is learning predictive knowledge about the environment in the form of value functions, from which it can derive its behavior to address long-term sequential decision making problems. The agent's horizon of interest, that is, how immediate or long-term a TD learning agent predicts into the future, is adjusted through a discount rate para…
Finite-Time Accuracy of Temporal-Difference Learning Under Schur-Stable Recursions
Department of Electrical Engineering, KAIST, Daejeon, 34141, South Korea（韩国釜山科学技术院电子工程系）；Department of Defense Systems Engineering, Sejong University, Seoul, 05006,, South Korea（全州大学国防系统工程系）
Here, we provide a supplementary material for Takayuki Osogami, "Uncorrected least-squares temporal difference with lambda-return," which appears in {\it Proceedings of the 34th AAAI Conference on Artificial Intelligence} (AAAI-20).
