Can Active Sampling Reduce Causal Confusion in Offline Reinforcement Learning?

Causal confusion is a phenomenon where an agent learns a policy that reflects imperfect spurious correlations in the data. Such a policy may falsely appear to be optimal during training if most of the training data contain such spurious correlations. This phenomenon is particularly pronounced in domains such as robotics, with potentially large gaps between the open- and closed-loop performance of an agent. In such settings, causally confused models may appear to perform well according to open-loop metrics during training but fail catastrophically when deployed in the real world. In this paper, we study causal confusion in offline reinforcement learning. We investigate whether selectively sampling appropriate points from a dataset of demonstrations may enable offline reinforcement learning agents to disambiguate the underlying causal mechanisms of the environment, alleviate causal confusion in offline reinforcement learning, and produce a safer model for deployment. To answer this question, we consider a set of tailored offline reinforcement learning datasets that exhibit causal ambiguity and assess the ability of active sampling techniques to reduce causal confusion at evaluation. We provide empirical evidence that uniform and active sampling techniques are able to consistently reduce causal confusion as training progresses and that active sampling is able to do so significantly more efficiently than uniform sampling.

ReLU to the Rescue: Improve Your On-Policy Actor-Critic with Positive Advantages

This paper introduces a novel method for enhancing the effectiveness of on-policy Deep Reinforcement Learning (DRL) algorithms. Three surprisingly simple modifications to the A3C algorithm: (1) processing advantage estimates through a ReLU function, (2) spectral normalization, and (3) dropout, serve to not only improve efficacy but also yield a ``cautious'' DRL algorithm. Where on-policy algorithms such as Proximal Policy Optimization (PPO) and Asynchronous Advantage Actor-Critic (A3C) do not explicitly account for cautious interaction with the environment, our method integrates caution in two critical ways: (1) by maximizing a lower bound on the value function plus a constant, thereby promoting a \textit{conservative value estimation}, and (2) by incorporating Thompson sampling for cautious exploration. In proving that our algorithm maximizes the lower bound, we also ground Regret Matching Policy Gradients (RMPG), a discrete-action on-policy method for multi-agent reinforcement learning. Our rigorous empirical evaluations across various benchmarks demonstrate our approach's improved performance against existing on-policy algorithms. This research represents a substantial step towards efficacious and cautious DRL algorithms, which are needed to unlock applications to complex, real-world problems.

La-MAML: Look-ahead Meta Learning for Continual Learning

The continual learning problem involves training models with limited capacity to perform well on a set of an unknown number of sequentially arriving tasks. While meta-learning shows great potential for reducing interference between old and new tasks, the current training procedures tend to be either slow or offline, and sensitive to many hyper-parameters. In this work, we propose Look-ahead MAML (La-MAML), a fast optimisation-based meta-learning algorithm for online-continual learning, aided by a small episodic memory. Our proposed modulation of per-parameter learning rates in our meta-learning update allows us to draw connections to prior work on hypergradients and meta-descent. This provides a more flexible and efficient way to mitigate catastrophic forgetting compared to conventional prior-based methods. La-MAML achieves performance superior to other replay-based, prior-based and meta-learning based approaches for continual learning on real-world visual classification benchmarks.

Unifying Variational Inference and PAC-Bayes for Supervised Learning that Scales

Neural Network based controllers hold enormous potential to learn complex, high-dimensional functions. However, they are prone to overfitting and unwarranted extrapolations. PAC Bayes is a generalized framework which is more resistant to overfitting and that yields performance bounds that hold with arbitrarily high probability even on the unjustified extrapolations. However, optimizing to learn such a function and a bound is intractable for complex tasks. In this work, we propose a method to simultaneously learn such a function and estimate performance bounds that scale organically to high-dimensions, non-linear environments without making any explicit assumptions about the environment. We build our approach on a parallel that we draw between the formulations called ELBO and PAC Bayes when the risk metric is negative log likelihood. Through our experiments on multiple high dimensional MuJoCo locomotion tasks, we validate the correctness of our theory, show its ability to generalize better, and investigate the factors that are important for its learning. The code for all the experiments is available at https://bit.ly/2qv0JjA.

Geometric Consistency for Self-Supervised End-to-End Visual Odometry

With the success of deep learning based approaches in tackling challenging problems in computer vision, a wide range of deep architectures have recently been proposed for the task of visual odometry (VO) estimation. Most of these proposed solutions rely on supervision, which requires the acquisition of precise ground-truth camera pose information, collected using expensive motion capture systems or high-precision IMU/GPS sensor rigs. In this work, we propose an unsupervised paradigm for deep visual odometry learning. We show that using a noisy teacher, which could be a standard VO pipeline, and by designing a loss term that enforces geometric consistency of the trajectory, we can train accurate deep models for VO that do not require ground-truth labels. We leverage geometry as a self-supervisory signal and propose "Composite Transformation Constraints (CTCs)", that automatically generate supervisory signals for training and enforce geometric consistency in the VO estimate. We also present a method of characterizing the uncertainty in VO estimates thus obtained. To evaluate our VO pipeline, we present exhaustive ablation studies that demonstrate the efficacy of end-to-end, self-supervised methodologies to train deep models for monocular VO. We show that leveraging concepts from geometry and incorporating them into the training of a recurrent neural network results in performance competitive to supervised deep VO methods.