This paper introduces new perspectives on analog design space search. To minimize the time-to-market, this endeavor better cast as constraint satisfaction problem than global optimization defined in prior arts. We incorporate model-based agents, contrasted with model-free learning, to implement a trust-region strategy. As such, simple feed-forward networks can be trained with supervised learning, where the convergence is relatively trivial. Experiment results demonstrate orders of magnitude improvement on search iterations. Additionally, the unprecedented consideration of PVT conditions are accommodated. On circuits with TSMC 5/6nm process, our method achieve performance surpassing human designers. Furthermore, this framework is in production in industrial settings.
Effective planning in model-based reinforcement learning (MBRL) and model-predictive control (MPC) relies on the accuracy of the learned dynamics model. In many instances of MBRL and MPC, this model is assumed to be stationary and is periodically re-trained from scratch on state transition experience collected from the beginning of environment interactions. This implies that the time required to train the dynamics model - and the pause required between plan executions - grows linearly with the size of the collected experience. We argue that this is too slow for lifelong robot learning and propose HyperCRL, a method that continually learns the encountered dynamics in a sequence of tasks using task-conditional hypernetworks. Our method has three main attributes: first, it includes dynamics learning sessions that do not revisit training data from previous tasks, so it only needs to store the most recent fixed-size portion of the state transition experience; second, it uses fixed-capacity hypernetworks to represent non-stationary and task-aware dynamics; third, it outperforms existing continual learning alternatives that rely on fixed-capacity networks, and does competitively with baselines that remember an ever increasing coreset of past experience. We show that HyperCRL is effective in continual model-based reinforcement learning in robot locomotion and manipulation scenarios, such as tasks involving pushing and door opening. Our project website with videos is at this link https://rvl.cs.toronto.edu/blog/2020/hypercrl
Existing model-based value expansion methods typically leverage a world model for value estimation with a fixed rollout horizon to assist policy learning. However, the fixed rollout with an inaccurate model has a potential to harm the learning process. In this paper, we investigate the idea of using the model knowledge for value expansion adaptively. We propose a novel method called Dynamic-horizon Model-based Value Expansion (DMVE) to adjust the world model usage with different rollout horizons. Inspired by reconstruction-based techniques that can be applied for visual data novelty detection, we utilize a world model with a reconstruction module for image feature extraction, in order to acquire more precise value estimation. The raw and the reconstructed images are both used to determine the appropriate horizon for adaptive value expansion. On several benchmark visual control tasks, experimental results show that DMVE outperforms all baselines in sample efficiency and final performance, indicating that DMVE can achieve more effective and accurate value estimation than state-of-the-art model-based methods.
Despite its experimental success, Model-based Reinforcement Learning still lacks a complete theoretical understanding. To this end, we analyze the error in the cumulative reward using a contraction approach. We consider both stochastic and deterministic state transitions for continuous (non-discrete) state and action spaces. This approach doesn't require strong assumptions and can recover the typical quadratic error to the horizon. We prove that branched rollouts can reduce this error and are essential for deterministic transitions to have a Bellman contraction. Our analysis of policy mismatch error also applies to Imitation Learning. In this case, we show that GAN-type learning has an advantage over Behavioral Cloning when its discriminator is well-trained.
Active inference is a probabilistic framework for modelling the behaviour of biological and artificial agents, which derives from the principle of minimising free energy. In recent years, this framework has successfully been applied to a variety of situations where the goal was to maximise reward, offering comparable and sometimes superior performance to alternative approaches. In this paper, we clarify the connection between reward maximisation and active inference by demonstrating how and when active inference agents perform actions that are optimal for maximising reward. Precisely, we show the conditions under which active inference produces the optimal solution to the Bellman equation--a formulation that underlies several approaches to model-based reinforcement learning and control. On partially observed Markov decision processes, the standard active inference scheme can produce Bellman optimal actions for planning horizons of 1, but not beyond. In contrast, a recently developed recursive active inference scheme (sophisticated inference) can produce Bellman optimal actions on any finite temporal horizon. We append the analysis with a discussion of the broader relationship between active inference and reinforcement learning.
Severe infectious diseases such as the novel coronavirus (COVID-19) pose a huge threat to public health. Stringent control measures, such as school closures and stay-at-home orders, while having significant effects, also bring huge economic losses. In the face of an emerging infectious disease, a crucial question for policymakers is how to make the trade-off and implement the appropriate interventions timely given the huge uncertainty. In this work, we propose a Multi-Objective Model-based Reinforcement Learning framework to facilitate data-driven decision-making and minimize the overall long-term cost. Specifically, at each decision point, a Bayesian epidemiological model is first learned as the environment model, and then the proposed model-based multi-objective planning algorithm is applied to find a set of Pareto-optimal policies. This framework, combined with the prediction bands for each policy, provides a real-time decision support tool for policymakers. The application is demonstrated with the spread of COVID-19 in China.
We present a novel approach (DyNODE) that captures the underlying dynamics of a system by incorporating control in a neural ordinary differential equation framework. We conduct a systematic evaluation and comparison of our method and standard neural network architectures for dynamics modeling. Our results indicate that a simple DyNODE architecture when combined with an actor-critic reinforcement learning (RL) algorithm that uses model predictions to improve the critic's target values, outperforms canonical neural networks, both in sample efficiency and predictive performance across a diverse range of continuous tasks that are frequently used to benchmark RL algorithms. This approach provides a new avenue for the development of models that are more suited to learn the evolution of dynamical systems, particularly useful in the context of model-based reinforcement learning. To assist related work, we have made code available at https://github.com/vmartinezalvarez/DyNODE .
For over a decade, model-based reinforcement learning has been seen as a way to leverage control-based domain knowledge to improve the sample-efficiency of reinforcement learning agents. While model-based agents are conceptually appealing, their policies tend to lag behind those of model-free agents in terms of final reward, especially in non-trivial environments. In response, researchers have proposed model-based agents with increasingly complex components, from ensembles of probabilistic dynamics models, to heuristics for mitigating model error. In a reversal of this trend, we show that simple model-based agents can be derived from existing ideas that not only match, but outperform state-of-the-art model-free agents in terms of both sample-efficiency and final reward. We find that a model-free soft value estimate for policy evaluation and a model-based stochastic value gradient for policy improvement is an effective combination, achieving state-of-the-art results on a high-dimensional humanoid control task, which most model-based agents are unable to solve. Our findings suggest that model-based policy evaluation deserves closer attention.
Intelligent assistants that follow commands or answer simple questions, such as Siri and Google search, are among the most economically important applications of AI. Future conversational AI assistants promise even greater capabilities and a better user experience through a deeper understanding of the domain, the user, or the user's purposes. But what domain and what methods are best suited to researching and realizing this promise? In this article we argue for the domain of voice document editing and for the methods of model-based reinforcement learning. The primary advantages of voice document editing are that the domain is tightly scoped and that it provides something for the conversation to be about (the document) that is delimited and fully accessible to the intelligent assistant. The advantages of reinforcement learning in general are that its methods are designed to learn from interaction without explicit instruction and that it formalizes the purposes of the assistant. Model-based reinforcement learning is needed in order to genuinely understand the domain of discourse and thereby work efficiently with the user to achieve their goals. Together, voice document editing and model-based reinforcement learning comprise a promising research direction for achieving conversational AI.
Trajectory optimizers for model-based reinforcement learning, such as the Cross-Entropy Method (CEM), can yield compelling results even in high-dimensional control tasks and sparse-reward environments. However, their sampling inefficiency prevents them from being used for real-time planning and control. We propose an improved version of the CEM algorithm for fast planning, with novel additions including temporally-correlated actions and memory, requiring 2.7-22x less samples and yielding a performance increase of 1.2-10x in high-dimensional control problems.
Despite deep neural network (DNN)'s impressive prediction performance in various domains, it is well known now that a set of DNN models trained with the same model specification and the same data can produce very different prediction results. Ensemble method is one state-of-the-art benchmark for prediction uncertainty estimation. However, ensembles are expensive to train and serve for web-scale traffic. In this paper, we seek to advance the understanding of prediction variation estimated by the ensemble method. Through empirical experiments on two widely used benchmark datasets MovieLens and Criteo in recommender systems, we observe that prediction variations come from various randomness sources, including training data shuffling, and parameter random initialization. By introducing more randomness into model training, we notice that ensemble's mean predictions tend to be more accurate while the prediction variations tend to be higher. Moreover, we propose to infer prediction variation from neuron activation strength and demonstrate the strong prediction power from activation strength features. Our experiment results show that the average R squared on MovieLens is as high as 0.56 and on Criteo is 0.81. Our method performs especially well when detecting the lowest and highest variation buckets, with 0.92 AUC and 0.89 AUC respectively. Our approach provides a simple way for prediction variation estimation, which opens up new opportunities for future work in many interesting areas (e.g.,model-based reinforcement learning) without relying on serving expensive ensemble models.
Deep reinforcement learning has shown remarkable success in the past few years. Highly complex sequential decision making problems have been solved in tasks such as game playing and robotics. Unfortunately, the sample complexity of most deep reinforcement learning methods is high, precluding their use in some important applications. Model-based reinforcement learning creates an explicit model of the environment dynamics to reduce the need for environment samples. Current deep learning methods use high-capacity networks to solve high-dimensional problems. Unfortunately, high-capacity models typically require many samples, negating the potential benefit of lower sample complexity in model-based methods. A challenge for deep model-based methods is therefore to achieve high predictive power while maintaining low sample complexity. In recent years, many model-based methods have been introduced to address this challenge. In this paper, we survey the contemporary model-based landscape. First we discuss definitions and relations to other fields. We propose a taxonomy based on three approaches: using explicit planning on given transitions, using explicit planning on learned transitions, and end-to-end learning of both planning and transitions. We use these approaches to organize a comprehensive overview of important recent developments such as latent models. We describe methods and benchmarks, and we suggest directions for future work for each of the approaches. Among promising research directions are curriculum learning, uncertainty modeling, and use of latent models for transfer learning.
This paper introduces an algorithm for discovering implicit and delayed causal relations between events observed by a robot at arbitrary times, with the objective of improving data-efficiency and interpretability of model-based reinforcement learning (RL) techniques. The proposed algorithm initially predicts observations with the Markov assumption, and incrementally introduces new hidden variables to explain and reduce the stochasticity of the observations. The hidden variables are memory units that keep track of pertinent past events. Such events are systematically identified by their information gains. The learned transition and reward models are then used for planning. Experiments on simulated and real robotic tasks show that this method significantly improves over current RL techniques.
In the present paper, we propose a decoder-free extension of Dreamer, a leading model-based reinforcement learning (MBRL) method from pixels. Dreamer is a sample- and cost-efficient solution to robot learning, as it is used to train latent state-space models based on a variational autoencoder and to conduct policy optimization by latent trajectory imagination. However, this autoencoding based approach often causes object vanishing, in which the autoencoder fails to perceives key objects for solving control tasks, and thus significantly limiting Dreamer's potential. This work aims to relieve this Dreamer's bottleneck and enhance its performance by means of removing the decoder. For this purpose, we firstly derive a likelihood-free and InfoMax objective of contrastive learning from the evidence lower bound of Dreamer. Secondly, we incorporate two components, (i) independent linear dynamics and (ii) the random crop data augmentation, to the learning scheme so as to improve the training performance. In comparison to Dreamer and other recent model-free reinforcement learning methods, our newly devised Dreamer with InfoMax and without generative decoder (Dreaming) achieves the best scores on 5 difficult simulated robotics tasks, in which Dreamer suffers from object vanishing.
In this work, we study model-based reinforcement learning (RL) in unknown stabilizable linear dynamical systems. When learning a dynamical system, one needs to stabilize the unknown dynamics in order to avoid system blow-ups. We propose an algorithm that certifies fast stabilization of the underlying system by effectively exploring the environment with an improved exploration strategy. We show that the proposed algorithm attains $\tilde{\mathcal{O}}(\sqrt{T})$ regret after $T$ time steps of agent-environment interaction. We also show that the regret of the proposed algorithm has only a polynomial dependence in the problem dimensions, which gives an exponential improvement over the prior methods. Our improved exploration method is simple, yet efficient, and it combines a sophisticated exploration policy in RL with an isotropic exploration strategy to achieve fast stabilization and improved regret. We empirically demonstrate that the proposed algorithm outperforms other popular methods in several adaptive control tasks.
Prioritized Experience Replay (ER) has been empirically shown to improve sample efficiency across many domains and attracted great attention; however, there is little theoretical understanding of why such prioritized sampling helps and its limitations. In this work, we take a deep look at the prioritized ER. In a supervised learning setting, we show the equivalence between the error-based prioritized sampling method for mean squared error and uniform sampling for cubic power loss. We then provide theoretical insight into why it improves convergence rate upon uniform sampling during early learning. Based on the insight, we further point out two limitations of the prioritized ER method: 1) outdated priorities and 2) insufficient coverage of the sample space. To mitigate the limitations, we propose our model-based stochastic gradient Langevin dynamics sampling method. We show that our method does provide states distributed close to an ideal prioritized sampling distribution estimated by the brute-force method, which does not suffer from the two limitations. We conduct experiments on both discrete and continuous control problems to show our approach's efficacy and examine the practical implication of our method in an autonomous driving application.
Model-based reinforcement learning (RL), which finds an optimal policy using an empirical model, has long been recognized as one of the corner stones of RL. It is especially suitable for multi-agent RL (MARL), as it naturally decouples the learning and the planning phases, and avoids the non-stationarity problem when all agents are improving their policies simultaneously using samples. Though intuitive and widely-used, the sample complexity of model-based MARL algorithms has not been fully investigated. In this paper, our goal is to address the fundamental question about its sample complexity. We study arguably the most basic MARL setting: two-player discounted zero-sum Markov games, given only access to a generative model. We show that model-based MARL achieves a sample complexity of $\tilde O(|S||A||B|(1-\gamma)^{-3}\epsilon^{-2})$ for finding the Nash equilibrium (NE) value up to some $\epsilon$ error, and the $\epsilon$-NE policies with a smooth planning oracle, where $\gamma$ is the discount factor, and $S,A,B$ denote the state space, and the action spaces for the two agents. We further show that such a sample bound is minimax-optimal (up to logarithmic factors) if the algorithm is reward-agnostic, where the algorithm queries state transition samples without reward knowledge, by establishing a matching lower bound. This is in contrast to the usual reward-aware setting, with a $\tilde\Omega(|S|(|A|+|B|)(1-\gamma)^{-3}\epsilon^{-2})$ lower bound, where this model-based approach is near-optimal with only a gap on the $|A|,|B|$ dependence. Our results not only demonstrate the sample-efficiency of this basic model-based approach in MARL, but also elaborate on the fundamental tradeoff between its power (easily handling the more challenging reward-agnostic case) and limitation (less adaptive and suboptimal in $|A|,|B|$), particularly arises in the multi-agent context.
Model-based reinforcement learning (RL) is appealing because (i) it enables planning and thus more strategic exploration, and (ii) by decoupling dynamics from rewards, it enables fast transfer to new reward functions. However, learning an accurate Markov Decision Process (MDP) over high-dimensional states (e.g., raw pixels) is extremely challenging because it requires function approximation, which leads to compounding errors. Instead, to avoid compounding errors, we propose learning an abstract MDP over abstract states: low-dimensional coarse representations of the state (e.g., capturing agent position, ignoring other objects). We assume access to an abstraction function that maps the concrete states to abstract states. In our approach, we construct an abstract MDP, which grows through strategic exploration via planning. Similar to hierarchical RL approaches, the abstract actions of the abstract MDP are backed by learned subpolicies that navigate between abstract states. Our approach achieves strong results on three of the hardest Arcade Learning Environment games (Montezuma's Revenge, Pitfall!, and Private Eye), including superhuman performance on Pitfall! without demonstrations. After training on one task, we can reuse the learned abstract MDP for new reward functions, achieving higher reward in 1000x fewer samples than model-free methods trained from scratch.
Deep model-based Reinforcement Learning (RL) has the potential to substantially improve the sample-efficiency of deep RL. While various challenges have long held it back, a number of papers have recently come out reporting success with deep model-based methods. This is a great development, but the lack of a consistent metric to evaluate such methods makes it difficult to compare various approaches. For example, the common single-task sample-efficiency metric conflates improvements due to model-based learning with various other aspects, such as representation learning, making it difficult to assess true progress on model-based RL. To address this, we introduce an experimental setup to evaluate model-based behavior of RL methods, inspired by work from neuroscience on detecting model-based behavior in humans and animals. Our metric based on this setup, the Local Change Adaptation (LoCA) regret, measures how quickly an RL method adapts to a local change in the environment. Our metric can identify model-based behavior, even if the method uses a poor representation and provides insight in how close a method's behavior is from optimal model-based behavior. We use our setup to evaluate the model-based behavior of MuZero on a variation of the classic Mountain Car task.
Model-based reinforcement learning approaches leverage a forward dynamics model to support planning and decision making, which, however, may fail catastrophically if the model is inaccurate. Although there are several existing methods dedicated to combating the model error, the potential of the single forward model is still limited. In this paper, we propose to additionally construct a backward dynamics model to reduce the reliance on accuracy in forward model predictions. We develop a novel method, called Bidirectional Model-based Policy Optimization (BMPO) to utilize both the forward model and backward model to generate short branched rollouts for policy optimization. Furthermore, we theoretically derive a tighter bound of return discrepancy, which shows the superiority of BMPO against the one using merely the forward model. Extensive experiments demonstrate that BMPO outperforms state-of-the-art model-based methods in terms of sample efficiency and asymptotic performance.