Recent advances in statistical learning theory have revealed profound connections between mutual information (MI) bounds, PAC-Bayesian theory, and Bayesian nonparametrics. This work introduces a novel mutual information bound for statistical models. The derived bound has wide-ranging applications in statistical inference. It yields improved contraction rates for fractional posteriors in Bayesian nonparametrics. It can also be used to study a wide range of estimation methods, such as variational inference or Maximum Likelihood Estimation (MLE). By bridging these diverse areas, this work advances our understanding of the fundamental limits of statistical inference and the role of information in learning from data. We hope that these results will not only clarify connections between statistical inference and information theory but also help to develop a new toolbox to study a wide range of estimators.

Gaussian graphical model selection is an important paradigm with numerous applications, including biological network modeling, financial network modeling, and social network analysis. Traditional approaches assume access to independent and identically distributed (i.i.d) samples, which is often impractical in real-world scenarios. In this paper, we address Gaussian graphical model selection under observations from a more realistic dependent stochastic process known as Glauber dynamics. Glauber dynamics, also called the Gibbs sampler, is a Markov chain that sequentially updates the variables of the underlying model based on the statistics of the remaining model. Such models, aside from frequently being employed to generate samples from complex multivariate distributions, naturally arise in various settings, such as opinion consensus in social networks and clearing/stock-price dynamics in financial networks. In contrast to the extensive body of existing work, we present the first algorithm for Gaussian graphical model selection when data are sampled according to the Glauber dynamics. We provide theoretical guarantees on the computational and statistical complexity of the proposed algorithm's structure learning performance. Additionally, we provide information-theoretic lower bounds on the statistical complexity and show that our algorithm is nearly minimax optimal for a broad class of problems.

This Dissertation is comprised of two main projects, addressing questions in neuroscience through applications of generative modeling. Project #1 (Chapter 4) explores how neurons encode features of the external world. I combine Helmholtz's "Perception as Unconscious Inference" -- paralleled by modern generative models like variational autoencoders (VAE) -- with the hierarchical structure of the visual cortex. This combination leads to the development of a hierarchical VAE model, which I test for its ability to mimic neurons from the primate visual cortex in response to motion stimuli. Results show that the hierarchical VAE perceives motion similar to the primate brain. Additionally, the model identifies causal factors of retinal motion inputs, such as object- and self-motion, in a completely unsupervised manner. Collectively, these results suggest that hierarchical inference underlines the brain's understanding of the world, and hierarchical VAEs can effectively model this understanding. Project #2 (Chapter 5) investigates the spatiotemporal structure of spontaneous brain activity and its reflection of brain states like rest. Using simultaneous fMRI and wide-field Ca2+ imaging data, this project demonstrates that the mouse cortex can be decomposed into overlapping communities, with around half of the cortical regions belonging to multiple communities. Comparisons reveal similarities and differences between networks inferred from fMRI and Ca2+ signals. The introduction (Chapter 1) is divided similarly to this abstract: sections 1.1 to 1.8 provide background information about Project #1, and sections 1.9 to 1.13 are related to Project #2. Chapter 2 includes historical background, Chapter 3 provides the necessary mathematical background, and finally, Chapter 6 contains concluding remarks and future directions.

Tensor Robust Principal Component Analysis (TRPCA) holds a crucial position in machine learning and computer vision. It aims to recover underlying low-rank structures and characterizing the sparse structures of noise. Current approaches often encounter difficulties in accurately capturing the low-rank properties of tensors and balancing the trade-off between low-rank and sparse components, especially in a mixed-noise scenario. To address these challenges, we introduce a Bayesian framework for TRPCA, which integrates a low-rank tensor nuclear norm prior and a generalized sparsity-inducing prior. By embedding the proposed priors within the Bayesian framework, our method can automatically determine the optimal tensor nuclear norm and achieve a balance between the nuclear norm and sparse components. Furthermore, our method can be efficiently extended to the weighted tensor nuclear norm model. Experiments conducted on synthetic and real-world datasets demonstrate the effectiveness and superiority of our method compared to state-of-the-art approaches.

Predictions in environments where a mix of legal policies, physical limitations, and operational preferences impacts an agent's motion are inherently difficult. Since Neuro-Symbolic systems allow for differentiable information flow between deep learning and symbolic building blocks, they present a promising avenue for expressing such high-level constraints. While prior work has demonstrated how to establish novel planning setups, e.g., in advanced aerial mobility tasks, their application in prediction tasks has been underdeveloped. We present the Constitutional Filter (CoFi), a novel filter architecture leveraging a Neuro-Symbolic representation of an agent's rules, i.e., its constitution, to (i) improve filter accuracy, (ii) leverage expert knowledge, (iii) incorporate deep learning architectures, and (iv) account for uncertainties in the environments through probabilistic spatial relations. CoFi follows a general, recursive Bayesian estimation setting, making it compatible with a vast landscape of estimation techniques such as Particle Filters. To underpin the advantages of CoFi, we validate its performance on real-world marine data from the Automatic Identification System and official Electronic Navigational Charts.

Recent work revealed a tight connection between adversarial robustness and restricted forms of symbolic explanations, namely distance-based (formal) explanations. This connection is significant because it represents a first step towards making the computation of symbolic explanations as efficient as deciding the existence of adversarial examples, especially for highly complex machine learning (ML) models. However, a major performance bottleneck remains, because of the very large number of features that ML models may possess, in particular for deep neural networks. This paper proposes novel algorithms to compute the so-called contrastive explanations for ML models with a large number of features, by leveraging on adversarial robustness. Furthermore, the paper also proposes novel algorithms for listing explanations and finding smallest contrastive explanations. The experimental results demonstrate the performance gains achieved by the novel algorithms proposed in this paper.

Given the ease of creating synthetic data from machine learning models, new models can be potentially trained on synthetic data generated by previous models. This recursive training process raises concerns about the long-term impact on model quality. As models are recursively trained on generated data from previous rounds, their ability to capture the nuances of the original human-generated data may degrade. This is often referred to as \emph{model collapse}. In this work, we ask how fast model collapse occurs for some well-studied distribution families under maximum likelihood (ML or near ML) estimation during recursive training. Surprisingly, even for fundamental distributions such as discrete and Gaussian distributions, the exact rate of model collapse is unknown. In this work, we theoretically characterize the rate of collapse in these fundamental settings and complement it with experimental evaluations. Our results show that for discrete distributions, the time to forget a word is approximately linearly dependent on the number of times it occurred in the original corpus, and for Gaussian models, the standard deviation reduces to zero roughly at $n$ iterations, where $n$ is the number of samples at each iteration. Both of these findings imply that model forgetting, at least in these simple distributions under near ML estimation with many samples, takes a long time.

Sickle cell anemia, which is characterized by abnormal erythrocyte morphology, can be detected using microscopic images. Computational techniques in medicine enhance the diagnosis and treatment efficiency. However, many computational techniques, particularly those based on Convolutional Neural Networks (CNNs), require high resources and time for training, highlighting the research opportunities in methods with low computational overhead. In this paper, we propose a novel approach combining conventional classifiers, segmented images, and CNNs for the automated classification of sickle cell disease. We evaluated the impact of segmented images on classification, providing insight into deep learning integration. Our results demonstrate that using segmented images and CNN features with an SVM achieves an accuracy of 96.80%. This finding is relevant for computationally efficient scenarios, paving the way for future research and advancements in medical-image analysis.

In this study, we introduce a novel methodological framework called Bayesian Penalized Empirical Likelihood (BPEL), designed to address the computational challenges inherent in empirical likelihood (EL) approaches. Our approach has two primary objectives: (i) to enhance the inherent flexibility of EL in accommodating diverse model conditions, and (ii) to facilitate the use of well-established Markov Chain Monte Carlo (MCMC) sampling schemes as a convenient alternative to the complex optimization typically required for statistical inference using EL. To achieve the first objective, we propose a penalized approach that regularizes the Lagrange multipliers, significantly reducing the dimensionality of the problem while accommodating a comprehensive set of model conditions. For the second objective, our study designs and thoroughly investigates two popular sampling schemes within the BPEL context. We demonstrate that the BPEL framework is highly flexible and efficient, enhancing the adaptability and practicality of EL methods. Our study highlights the practical advantages of using sampling techniques over traditional optimization methods for EL problems, showing rapid convergence to the global optima of posterior distributions and ensuring the effective resolution of complex statistical inference challenges.

Sampling from probability distributions in high dimensional spaces is a fundamental computational primitive; it forms the basis of efficient numerical methods for approximating arbitrary integrals. The problem statement is the following: given a density function π, compute a point x with density proportional to π(x). A general approach to solving this problem is to design a reversible, ergodic Markov chain with a unique stationary distribution that is equal to the target distribution from which samples are needed. It is often possible to design relatively simple chains with low per-iteration computational complexity that are fit for purpose by implementing the Metropolis-Hastings filter [1, 2], a rule by which to either accept the next step in the dynamics or remain put and so tailor the dynamics toward a specific stationary distribution. The resulting Metropolized or Markov Chain Monte Carlo algorithms are known to converge asymptotically to their stationary distributions under mild regularity conditions. Non-asymptotic rates of convergence or mixing times are comparatively few in number and are both algorithm-and target-specific. They are important because downstream estimators computed using samples drawn from a dynamics that has not converged will suffer from bias. The class of log-concave target distributions is of particular interest.