skip to main content


Title: Learning Value Functions from Undirected State-only Experience
Abstract: This paper tackles the problem of learning value functions from undirected state-only experience (state transitions without action labels i.e. (s,s’,r) tuples). We first theoretically characterize the applicability of Q-learning in this setting. We show that tabular Q-learning in discrete Markov decision processes (MDPs) learns the same value function under any arbitrary refinement of the action space. This theoretical result motivates the design of Latent Action Q-learning or LAQ, an offline RL method that can learn effective value functions from state-only experience. Latent Action Q-learning (LAQ) learns value functions using Q-learning on discrete latent actions obtained through a latent-variable future prediction model. We show that LAQ can recover value functions that have high correlation with value functions learned using ground truth actions. Value functions learned using LAQ lead to sample efficient acquisition of goal-directed behavior, can be used with domain-specific low-level controllers, and facilitate transfer across embodiments. Our experiments in 5 environments ranging from 2D grid world to 3D visual navigation in realistic environments demonstrate the benefits of LAQ over simpler alternatives, imitation learning oracles, and competing methods.  more » « less
Award ID(s):
2007035
NSF-PAR ID:
10416324
Author(s) / Creator(s):
; ;
Date Published:
Journal Name:
International Conference on Learning Representations (ICLR)
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. For many forms of e-learning environments, the system's behaviors can be viewed as a sequential decision process wherein, at each discrete step, the system is responsible for deciding the next system action when there are multiple ones available. Each of these system decisions a ects the user's successive actions and performance and some of them are more important than others. Thus, this raises an open ques- tion: how can we identify the critical system interactive de- cisions that are linked to student learning from a long trajec- tory of decisions? In this work, we proposed and evaluated Critical-Reinforcement Learning (Critical-RL), an adversar- ial deep reinforcement learning (ADRL) based framework to identify critical decisions and induce compact yet e ective policies. Speci cally, it induces a pair of adversarial policies based upon Deep Q-Network (DQN) with opposite goals: one is to improve student learning while the other is to hin- der; critical decisions are identi ed by comparing the two adversarial policies and using their corresponding Q-value di erences; nally, a Critical policy is induced by giving op- timal action on critical decisions but random yet reason- able decisions on others. We evaluated the e ectiveness of Critical policy against a random yet reasonable (Random) policy. While no signi cant di erence was found between the two condition, it is probably because of small sample sizes. Much to our surprise, we found that students often experience so-called Critical phase: a consecutive sequence of critical decisions with the same action. Students were further divided into High vs. Low based on the number of Critical phases they experienced and our results showed that while no signi cant was found between the two Low groups, the High Critical group learned signi cantly more than the High Random group. 
    more » « less
  2. Sequential decision-making under uncertainty is present in many important problems. Two popular approaches for tackling such problems are reinforcement learning and online search (e.g., Monte Carlo tree search). While the former learns a policy by interacting with the environment (typically done before execution), the latter uses a generative model of the environment to sample promising action trajectories at decision time. Decision-making is particularly challenging in non-stationary environments, where the environment in which an agent operates can change over time. Both approaches have shortcomings in such settings -- on the one hand, policies learned before execution become stale when the environment changes and relearning takes both time and computational effort. Online search, on the other hand, can return sub-optimal actions when there are limitations on allowed runtime. In this paper, we introduce \textit{Policy-Augmented Monte Carlo tree search} (PA-MCTS), which combines action-value estimates from an out-of-date policy with an online search using an up-to-date model of the environment. We prove theoretical results showing conditions under which PA-MCTS selects the one-step optimal action and also bound the error accrued while following PA-MCTS as a policy. We compare and contrast our approach with AlphaZero, another hybrid planning approach, and Deep Q Learning on several OpenAI Gym environments. Through extensive experiments, we show that under non-stationary settings with limited time constraints, PA-MCTS outperforms these baselines. 
    more » « less
  3. The practicality of reinforcement learning algorithms has been limited due to poor scaling with respect to the problem size, as the sample complexity of learning an ε-optimal policy is Ω(|S||A|H/ ε2) over worst case instances of an MDP with state space S, action space A, and horizon H. We consider a class of MDPs for which the associated optimal Q* function is low rank, where the latent features are unknown. While one would hope to achieve linear sample complexity in |S| and |A| due to the low rank structure, we show that without imposing further assumptions beyond low rank of Q*, if one is constrained to estimate the Q function using only observations from a subset of entries, there is a worst case instance in which one must incur a sample complexity exponential in the horizon H to learn a near optimal policy. We subsequently show that under stronger low rank structural assumptions, given access to a generative model, Low Rank Monte Carlo Policy Iteration (LR-MCPI) and Low Rank Empirical Value Iteration (LR-EVI) achieve the desired sample complexity of Õ((|S|+|A|)poly (d,H)/ε2) for a rank d setting, which is minimax optimal with respect to the scaling of |S|, |A|, and ε. In contrast to literature on linear and low-rank MDPs, we do not require a known feature mapping, our algorithm is computationally simple, and our results hold for long time horizons. Our results provide insights on the minimal low-rank structural assumptions required on the MDP with respect to the transition kernel versus the optimal action-value function. 
    more » « less
  4. We develop an approach to improve the learning capabilities of robotic systems by combining learned predictive models with experience-based state-action policy mappings. Predictive models provide an understanding of the task and the dynamics, while experience-based (model-free) policy mappings encode favorable actions that override planned actions. We refer to our approach of systematically combining model-based and model-free learning methods as hybrid learning. Our approach efficiently learns motor skills and improves the performance of predictive models and experience-based policies. Moreover, our approach enables policies (both model-based and model-free) to be updated using any off-policy reinforcement learning method. We derive a deterministic method of hybrid learning by optimally switching between learning modalities. We adapt our method to a stochastic variation that relaxes some of the key assumptions in the original derivation. Our deterministic and stochastic variations are tested on a variety of robot control benchmark tasks in simulation as well as a hardware manipulation task. We extend our approach for use with imitation learning methods, where experience is provided through demonstrations, and we test the expanded capability with a real-world pick-and-place task. The results show that our method is capable of improving the performance and sample efficiency of learning motor skills in a variety of experimental domains. 
    more » « less
  5. We present Q-functionals, an alternative architecture for continuous control deep reinforcement learning. Instead of returning a single value for a state-action pair, our network transforms a state into a function that can be rapidly evaluated in parallel for many actions, allowing us to efficiently choose high-value actions through sampling. This contrasts with the typical architecture of off-policy continuous control, where a policy network is trained for the sole purpose of selecting actions from the Q-function. We represent our action-dependent Q-function as a weighted sum of basis functions (Fourier, Polynomial, etc) over the action space, where the weights are state-dependent and output by the Q-functional network. Fast sampling makes practical a variety of techniques that require Monte-Carlo integration over Q-functions, and enables action-selection strategies besides simple value-maximization. We characterize our framework, describe various implementations of Q-functionals, and demonstrate strong performance on a suite of continuous control tasks. 
    more » « less