We provide the supplements of "Contextual Gaussian Process Bandits with Neural Networks" here. Specifically, we discuss alternative acquisition functions that can be incorporated with the neural network-accompanied Gaussian process (NN-AGP) model in Section 6. In Section 7, we discuss the bandit algorithm with NN-AGP, where the neural network approximation error is considered. In Section 8, we provide the detailed proof of theorems. We provide the experimental details and include additional numerical experiments in Section 9. Last we discuss the limitations of NN-AGP and propose the potential approaches to addressing the limitations for future work, including sparse NN-AGP for alleviating computational burdens and transfer learning with NN-AGP to address cold-start issue; see Section 10. In the main text, we employ the upper confidence bound function as the acquisition function in the contextual Bayesian optimization approach. Here, we provide two alternative choices: Thompson sampling (TS) and knowledge gradient (KG). We describe the two procedures of the contextual GP bandit problems with NN-AGP, where the acquisition function is replaced by TS or KG. It chooses the action that maximizes the expected reward with respect to a random belief that is drawn for a posterior distribution. Besides the multi-armed bandit problems, TS has also achieved both theoretical and practical success in BO and Gaussian process regression. For more detailed discussions on TS, we refer to [87, 88]. Specifically, we propose a neural network-accompanied Gaussian process Thompson sampling (NNAGP-TS) approach to address contextual GP bandits. The approach works as follows. In each iteration, NN-AGP-TS first fits an NN-AGP model with the historic data. Then, given the current contextual variable, a realization of the Gaussian process with respect to x X is sampled from the posterior distribution conditional on the historic data1.

Federated learning aims to learn a global model that performs well on client devices with limited cross-client communication. Personalized federated learning (PFL) further extends this setup to handle data heterogeneity between clients by learning personalized models. A key challenge in this setting is to learn effectively across clients even though each client has unique data that is often limited in size. Here we present pFedGP, a solution to PFL that is based on Gaussian processes (GPs) with deep kernel learning. GPs are highly expressive models that work well in the low data regime due to their Bayesian nature.

With a principled representation of uncertainty and closed form posterior updates, Gaussian processes (GPs) are a natural choice for online decision making. However, Gaussian processes typically require at least O(n2) computations for n training points, limiting their general applicability. Stochastic variational Gaussian processes (SVGPs) can provide scalable inference for a dataset of fixed size, but are difficult to efficiently condition on new data. We propose online variational conditioning (OVC), a procedure for efficiently conditioning SVGPs in an online setting that does not require re-training through the evidence lower bound with the addition of new data. OVC enables the pairing of SVGPs with advanced lookahead acquisition functions for black-box optimization, even with non-Gaussian likelihoods. We show OVC provides compelling performance in a range of applications including active learning of malaria incidence, and reinforcement learning on MuJoCo simulated robotic control tasks.

Decoders built on Gaussian processes (GPs) are enticing due to the marginalisation over the non-linear function space. Such models (also known as GP-LVMs) are often expensive and notoriously difficult to train in practice, but can be scaled using variational inference and inducing points. In this paper, we revisit active set approximations. We develop a new stochastic estimate of the log-marginal likelihood based on recently discovered links to cross-validation, and we propose a computationally efficient approximation thereof. We demonstrate that the resulting stochastic active sets (SAS) approximation significantly improves the robustness of GP decoder training, while reducing computational cost. The SAS-GP obtains more structure in the latent space, scales to many datapoints, and learns better representations than variational autoencoders, which is rarely the case for GP decoders.

We study model-free reinforcement learning (RL) in non-stationary finite-horizon episodic Markov decision processes (MDPs) without prior knowledge of the non-stationarity. We focus on the piecewise-stationary (PS) setting, where both the reward and transition dynamics can change an arbitrary number of times. We propose Detection Augmented Reinforcement Learning (DARLING), a modular wrapper for PS-RL that applies to both tabular and linear MDPs, without knowledge of the changes. Under certain change-point separation and reachability conditions, DARLING improves the best available dynamic regret bounds in both settings and yields strong empirical performance. We further establish the first minimax lower bounds for PS-RL in tabular and linear MDPs, showing that DARLING is the first nearly optimal algorithm. Experiments on standard benchmarks demonstrate that DARLING consistently surpasses the state-of-the-art methods across diverse non-stationary scenarios.

Good sparse approximations are essential for practical inference in Gaussian Processes as the computational cost of exact methods is prohibitive for large datasets. The Fully Independent Training Conditional (FITC) and the Variational Free Energy (VFE) approximations are two recent popular methods. Despite superficial similarities, these approximations have surprisingly different theoretical properties and behave differently in practice. We thoroughly investigate the two methods for regression both analytically and through illustrative examples, and draw conclusions to guide practical application.

Blind source separation (BSS) seeks to recover latent source signals from observed mixtures. Variational autoencoders (VAEs) offer a natural perspective for this problem: the latent variables can be interpreted as source components, the encoder can be viewed as a demixing mapping from observations to sources, and the decoder can be regarded as a remixing process from inferred sources back to observations. In this work, we propose AR-Flow VAE, a novel VAE-based framework for BSS in which each latent source is endowed with a parameter-adaptive autoregressive flow prior. This prior significantly enhances the flexibility of latent source modeling, enabling the framework to capture complex non-Gaussian behaviors and structured dependencies, such as temporal correlations, that are difficult to represent with conventional priors. In addition, the structured prior design assigns distinct priors to different latent dimensions, thereby encouraging the latent components to separate into different source signals under heterogeneous prior constraints. Experimental results validate the effectiveness of the proposed architecture for blind source separation. More importantly, this work provides a foundation for future investigations into the identifiability and interpretability of AR-Flow VAE.