We study the Non-Stationary Reinforcement Learning (RL) under distribution shifts in both finite-horizon episodic and infinite-horizon discounted Markov Decision Processes (MDPs). In the finite-horizon case, the transition functions may suddenly change at a particular episode. In the infinite-horizon setting, such changes can occur at an arbitrary time step during the agent's interaction with the environment. While the Q-learning Upper Confidence Bound algorithm (QUCB) can discover a proper policy during learning, due to the distribution shifts, this policy can exploit sub-optimal rewards after the shift happens. To address this issue, we propose Density-QUCB (DQUCB), a shift-aware Q-learning~UCB algorithm, which uses a transition density function to detect distribution shifts, then leverages its likelihood to enhance the uncertainty estimation quality of Q-learning~UCB, resulting in a balance between exploration and exploitation. Theoretically, we prove that our oracle DQUCB achieves a better regret guarantee than QUCB. Empirically, our DQUCB enjoys the computational efficiency of model-free RL and outperforms QUCB baselines by having a lower regret across RL tasks, as well as a real-world COVID-19 patient hospital allocation task using a Deep-Q-learning architecture.

By leveraging the representation power of deep neural networks, neural upper confidence bound (UCB) algorithms have shown success in contextual bandits. To further balance the exploration and exploitation, we propose Neural-$\sigma^2$-LinearUCB, a variance-aware algorithm that utilizes $\sigma^2_t$, i.e., an upper bound of the reward noise variance at round $t$, to enhance the uncertainty quantification quality of the UCB, resulting in a regret performance improvement. We provide an oracle version for our algorithm characterized by an oracle variance upper bound $\sigma^2_t$ and a practical version with a novel estimation for this variance bound. Theoretically, we provide rigorous regret analysis for both versions and prove that our oracle algorithm achieves a better regret guarantee than other neural-UCB algorithms in the neural contextual bandits setting. Empirically, our practical method enjoys a similar computational efficiency, while outperforming state-of-the-art techniques by having a better calibration and lower regret across multiple standard settings, including on the synthetic, UCI, MNIST, and CIFAR-10 datasets.

Morden deep ensembles technique achieves strong uncertainty estimation performance by going through multiple forward passes with different models. This is at the price of a high storage space and a slow speed in the inference (test) time. To address this issue, we propose Density-Regression, a method that leverages the density function in uncertainty estimation and achieves fast inference by a single forward pass. We prove it is distance aware on the feature space, which is a necessary condition for a neural network to produce high-quality uncertainty estimation under distribution shifts. Empirically, we conduct experiments on regression tasks with the cubic toy dataset, benchmark UCI, weather forecast with time series, and depth estimation under real-world shifted applications. We show that Density-Regression has competitive uncertainty estimation performance under distribution shifts with modern deep regressors while using a lower model size and a faster inference speed.

Prevalent deterministic deep-learning models suffer from significant over-confidence under distribution shifts. Probabilistic approaches can reduce this problem but struggle with computational efficiency. In this paper, we propose Density-Softmax, a fast and lightweight deterministic method to improve calibrated uncertainty estimation via a combination of density function with the softmax layer. By using the latent representation's likelihood value, our approach produces more uncertain predictions when test samples are distant from the training samples. Theoretically, we show that Density-Softmax can produce high-quality uncertainty estimation with neural networks, as it is the solution of minimax uncertainty risk and is distance-aware, thus reducing the over-confidence of the standard softmax. Empirically, our method enjoys similar computational efficiency as a single forward pass deterministic with standard softmax on the shifted toy, vision, and language datasets across modern deep-learning architectures. Notably, Density-Softmax uses 4 times fewer parameters than Deep Ensembles and 6 times lower latency than Rank-1 Bayesian Neural Network, while obtaining competitive predictive performance and lower calibration errors under distribution shifts.

Self-Supervised Learning (SSL) is crucial for real-world applications, especially in data-hungry domains such as healthcare and self-driving cars. In addition to a lack of labeled data, these applications also suffer from distributional shifts. Therefore, an SSL method should provide robust generalization and uncertainty estimation in the test dataset to be considered a reliable model in such high-stakes domains. However, existing approaches often focus on generalization, without evaluating the model's uncertainty. The ability to compare SSL techniques for improving these estimates is therefore critical for research on the reliability of self-supervision models. In this paper, we explore variants of SSL methods, including Jigsaw Puzzles, Context, Rotation, Geometric Transformations Prediction for vision, as well as BERT and GPT for language tasks. We train SSL in auxiliary learning for vision and pre-training for language model, then evaluate the generalization (in-out classification accuracy) and uncertainty (expected calibration error) across different distribution covariate shift datasets, including MNIST-C, CIFAR-10-C, CIFAR-10.1, and MNLI. Our goal is to create a benchmark with outputs from experiments, providing a starting point for new SSL methods in Reliable Machine Learning. All source code to reproduce results is available at https://github.com/hamanhbui/reliable_ssl_baselines.