arXiv research

Study explores learning behavior of GFlowNets, revealing key mechanisms.

The study initiates a theoretical analysis of dynamic benchmarking models.

The VAE's reconstruction ability is studied using PAC-Bayes theory.

We study the task of online boosting--combining online weak learners into an online strong learner. While batch boosting has a sound theoretical foundation, online boosting deserves more study from the theoretical perspective. In this paper, we carefully compare the differences between online and batch boosting, and pr…

In three-dimensional computational topology, the theory of normal surfaces is a tool of great theoretical and practical significance. Although this theory typically leads to exponential time algorithms, very little is known about how these algorithms perform in "typical" scenarios, or how far the best known theoretical…

We conjecture a Verlinde type formula for the moduli space of Higgs sheaves on a surface with a holomorphic 2-form. The conjecture specializes to a Verlinde formula for the moduli space of sheaves. Our formula interpolates between $K$-theoretic Donaldson invariants studied by the first named author and Nakajima-Yoshiok…

This paper explains the theoretical inductive bias of Isolation Forest.

Boosting strategies for merging vs. ensembling studies analyzed.

Theoretical model for iterative user discovery in recommender systems.

Paper studies the theoretical equivalence between implicit and explicit neural networks in high dimensions.

Study reveals mutual information is crucial for understanding algorithm performance in stochastic convex optimization.

Theoretical study of random forests for nonlinear time series.

Dimension reduction and variable selection are performed routinely in case-control studies, but the literature on the theoretical aspects of the resulting estimates is scarce. We bring our contribution to this literature by studying estimators obtained via L1 penalized likelihood optimization. We show that the optimize…

Stochastic volatility models describe asset prices $S_t$ as driven by an unobserved process capturing the random dynamics of volatility $σ_t$. Here, we quantify how much information about $σ_t$ can be inferred from asset prices $S_t$ in terms of Shannon's mutual information $I(S_t : σ_t)$. This motivates a careful nume…

Theoretical analysis of the error landscape of deep neural networks has garnered significant interest in recent years. In this work, we theoretically study the importance of noise in the trajectories of gradient descent towards optimal solutions in multi-layer neural networks. We show that adding noise (in different wa…

In this paper we study the learnability of deep random networks from both theoretical and practical points of view. On the theoretical front, we show that the learnability of random deep networks with sign activation drops exponentially with its depth. On the practical front, we find that the learnability drops sharply…

The paper studies properties of group relations induced by compatible coarse structures.

The goal of this paper is not to introduce a single algorithm or method, but to make theoretical steps towards fully understanding the training dynamics of generative adversarial networks. In order to substantiate our theoretical analysis, we perform targeted experiments to verify our assumptions, illustrate our claims…

Survey on multiplayer bandits, highlighting theoretical gaps and future directions.

Study gauge theory of real and quaternionic parabolic bundles over real curves.

Theoretical guarantees for neural estimators in parametric statistics are derived.

This study connects Jacobian regularization to adversarial robustness and improves generalization.

The paper explores game-theoretic alignment of LLMs with human preferences, finding limitations and conditions.

This paper provides theoretical guarantees for SPCA using the Elastic Net.

This article presents a theoretical model for a dynamic system based on sustainable development. Due to the relatively absence of theoretical studies and practical issues in the area of sustainable development, Romania aspires to the principles of sustainable development. Based on the concept as a process in which econ…

In this work, we study generalized entropies and information geometry in a group-theoretical framework. We explore the conditions that ensure the existence of some natural properties and at the same time of a group-theoretical structure for a large class of entropies. In addition, a method for defining new entropies, u…

This work analyzes statistical properties of SAM, showing it outperforms GD.

Theoretical study on AI models' resilience to data contamination during recursive training.

New L2 regularization improves softmax MAB performance.

Transfer learning has been proven effective when within-target labeled data is scarce. A lot of works have developed successful algorithms and empirically observed positive transfer effect that improves target generalization error using source knowledge. However, theoretical analysis of transfer learning is more challe…

Study shows neural network parameters converge to ridgelet spectrum.

The paper explores the information-theoretic nature of excess risk in machine learning.

The paper explores dual learning, a technique that improves machine translation and image transformation.

Neural ODEs provide a framework for studying the training dynamics of neural networks.

Study optimizes fairness in predictive models by balancing utility and separation.

New framework for discrete-state diffusion models reduces sample complexity.

Machine learning is often used in competitive scenarios: Participants learn and fit static models, and those models compete in a shared platform. The common assumption is that in order to win a competition one has to have the best predictive model, i.e., the model with the smallest out-sample error. Is that necessarily…

In this thesis, we study value distribution theoretical properties of the Gauss map of pseudo-algebraic minimal surfaces in n-dimensional Euclidean space. After reviewing basic facts, we give estimates for the number of exceptional values and the totally ramified value numbers and the corresponding unicity theorems for…

Improved VB algorithm for high-dimensional logistic regression with theoretical guarantees.

Theoretical study on how model architecture affects contrastive learning performance.

There is emerging interest in performing regression between distributions. In contrast to prediction on single instances, these machine learning methods can be useful for population-based studies or on problems that are inherently statistical in nature. The recently proposed distribution regression network (DRN) has sh…

We review a braid theoretic self-linking number formula and study its applications.

We analyze oversquashing in topological message-passing using relational structures.

Study on detecting and recovering hidden dense cycles in random graphs.

Proposes a new theoretical framework for PbRL that requires less human feedback.

Empirical study shows consistent meta-RL algorithms adapt to OOD tasks.

Study robust mean estimation under coordinate-level corruptions using Hamming distance.

New research shows existing information-theoretic methods can't establish minimax rates for gradient descent in stochastic convex optimization.