Xi Chen

Mianchu Wang, Rui Yang, Xi Chen et al.

Offline Goal-Conditioned RL (GCRL) offers a feasible paradigm for learning general-purpose policies from diverse and multi-task offline datasets. Despite notable recent progress, the predominant offline GCRL methods, mainly model-free, face constraints in handling limited data and generalizing to unseen goals. In this work, we propose Goal-conditioned Offline Planning (GOPlan), a novel model-based framework that contains two key phases: (1) pretraining a prior policy capable of capturing multi-modal action distribution within the multi-goal dataset; (2) employing the reanalysis method with planning to generate imagined trajectories for funetuning policies. Specifically, we base the prior policy on an advantage-weighted conditioned generative adversarial network, which facilitates distinct mode separation, mitigating the pitfalls of out-of-distribution (OOD) actions. For further policy optimization, the reanalysis method generates high-quality imaginary data by planning with learned models for both intra-trajectory and inter-trajectory goals. With thorough experimental evaluations, we demonstrate that GOPlan achieves state-of-the-art performance on various offline multi-goal navigation and manipulation tasks. Moreover, our results highlight the superior ability of GOPlan to handle small data budgets and generalize to OOD goals.

Xi Chen, Caylin Hickey

Deep reinforcement learning (deep RL) has achieved superior performance in complex sequential tasks by learning directly from image input. A deep neural network is used as a function approximator and requires no specific state information. However, one drawback of using only images as input is that this approach requires a prohibitively large amount of training time and data for the model to learn the state feature representation and approach reasonable performance. This is not feasible in real-world applications, especially when the data are expansive and training phase could introduce disasters that affect human safety. In this work, we use a human demonstration approach to speed up training for learning features and use the resulting pre-trained model to replace the neural network in the deep RL Deep Q-Network (DQN), followed by human interaction to further refine the model. We empirically evaluate our approach by using only a human demonstration model and modified DQN with human demonstration model included in the Microsoft AirSim car simulator. Our results show that (1) pre-training with human demonstration in a supervised learning approach is better and much faster at discovering features than DQN alone, (2) initializing the DQN with a pre-trained model provides a significant improvement in training time and performance even with limited human demonstration, and (3) providing the ability for humans to supply suggestions during DQN training can speed up the network's convergence on an optimal policy, as well as allow it to learn more complex policies that are harder to discover by random exploration.

Tim Salimans, Jonathan Ho, Xi Chen et al.

We explore the use of Evolution Strategies (ES), a class of black box optimization algorithms, as an alternative to popular MDP-based RL techniques such as Q-learning and Policy Gradients. Experiments on MuJoCo and Atari show that ES is a viable solution strategy that scales extremely well with the number of CPUs available: By using a novel communication strategy based on common random numbers, our ES implementation only needs to communicate scalars, making it possible to scale to over a thousand parallel workers. This allows us to solve 3D humanoid walking in 10 minutes and obtain competitive results on most Atari games after one hour of training. In addition, we highlight several advantages of ES as a black box optimization technique: it is invariant to action frequency and delayed rewards, tolerant of extremely long horizons, and does not need temporal discounting or value function approximation.

Xi Chen, Yan Duan, Rein Houthooft et al.

This paper describes InfoGAN, an information-theoretic extension to the Generative Adversarial Network that is able to learn disentangled representations in a completely unsupervised manner. InfoGAN is a generative adversarial network that also maximizes the mutual information between a small subset of the latent variables and the observation. We derive a lower bound to the mutual information objective that can be optimized efficiently, and show that our training procedure can be interpreted as a variation of the Wake-Sleep algorithm. Specifically, InfoGAN successfully disentangles writing styles from digit shapes on the MNIST dataset, pose from lighting of 3D rendered images, and background digits from the central digit on the SVHN dataset. It also discovers visual concepts that include hair styles, presence/absence of eyeglasses, and emotions on the CelebA face dataset. Experiments show that InfoGAN learns interpretable representations that are competitive with representations learned by existing fully supervised methods.

Yuchen Zhang, Xi Chen, Dengyong Zhou et al.

Crowdsourcing is a popular paradigm for effectively collecting labels at low cost. The Dawid-Skene estimator has been widely used for inferring the true labels from the noisy labels provided by non-expert crowdsourcing workers. However, since the estimator maximizes a non-convex log-likelihood function, it is hard to theoretically justify its performance. In this paper, we propose a two-stage efficient algorithm for multi-class crowd labeling problems. The first stage uses the spectral method to obtain an initial estimate of parameters. Then the second stage refines the estimation by optimizing the objective function of the Dawid-Skene estimator via the EM algorithm. We show that our algorithm achieves the optimal convergence rate up to a logarithmic factor. We conduct extensive experiments on synthetic and real datasets. Experimental results demonstrate that the proposed algorithm is comparable to the most accurate empirical approach, while outperforming several other recently proposed methods.