arXiv research

Survey of financial foundation models for diverse applications.

GAMformer bridges tabular models and interpretability, offering a single-pass approach.

Study shows cliff-learning in transfer learning from foundation models.

Distance metric learning is a branch of machine learning that aims to learn distances from the data, which enhances the performance of similarity-based algorithms. This tutorial provides a theoretical background and foundations on this topic and a comprehensive experimental analysis of the most-known algorithms. We sta…

This article gives the foundations of the colored Jones polynomial for singular knots. We extend Masbum and Vogel's algorithm to compute the colored Jones polynomial for any singular knot. We also introduce the tail of the colored Jones polynomial of singular knots and use its stability properties to prove a false thet…

Foundation models improve wage gap decomposition by capturing omitted career history factors.

We develop the mathematical foundations of the stochastic modified equations (SME) framework for analyzing the dynamics of stochastic gradient algorithms, where the latter is approximated by a class of stochastic differential equations with small noise parameters. We prove that this approximation can be understood math…

Framework tackles class imbalance and noisy labels in active learning.

Survey of statistical queries and their applications.

New framework for optimizing machine learning risks.

Introduces foundation priors for using model-generated data in empirical research.

Revises Bayesian model averaging for foundation models.

The study provides theoretical foundations for using smaller instances to predict algorithm performance on larger ones.

Time series foundation models are well-calibrated, improving over baseline models.

Combines foundation models with weak supervision to improve NLP and video tasks.

Foundation models trained on collider data improve jet generation tasks.

New measure assesses time series pre-training data quality without labels.

Disagreement-based approaches generate multiple classifiers and exploit the disagreement among them with unlabeled data to improve learning performance. Co-training is a representative paradigm of them, which trains two classifiers separately on two sufficient and redundant views; while for the applications where there…

DAG-FM discovers causal relationships from heterogeneous data.

New method extracts biological concepts from cell microscopy images.

This thesis explores robust machine learning against adversarial examples.

Paper explores foundation models for dynamical systems using synthetic data.

A model learns causal graphs from summary statistics of synthetic data.

Paper speeds up large foundation models for time series data.

We show that a large class of Estimation of Distribution Algorithms, including, but not limited to, Covariance Matrix Adaption, can be written as a Monte Carlo Expectation-Maximization algorithm, and as exact EM in the limit of infinite samples. Because EM sits on a rigorous statistical foundation and has been thorough…

New method reduces fine-tuning cost for reused models.

Foundation models outperform supervised methods in time series forecasting across various operational regimes.

Improved regret bound for linear ensemble sampling.

Paper proposes verifier engineering for improving foundation models.

Foundation models improve time series prediction reliability, especially with limited data.

Foundation models improve on econometric benchmarks for forecasting volatility, but vary widely across models.

Tabular foundation models outperform other methods in conditional density estimation across various datasets.

Stabilizing black-box algorithms through task-oriented randomization

Foundation AI model outperforms traditional VaR methods in forecasting.

Mantis improves time series classification using a transformer model trained on synthetic data.

Paper introduces TS-GPT for engineering time series forecasting.

Framework fine-tunes foundation models with semi-supervised learning for downstream tasks and latent spaces.

Study evaluates interpretability of time series foundation models' latent spaces.

RTE enables extrapolation to new tasks by learning task transformations.

ML reduces high-dimensional data to reveal its underlying structure.

New framework explains leading digit patterns without probabilistic assumptions.

Current XAI research lacks solid foundations and clear goals.

VMoER improves uncertainty quantification in MoE layers for scalable foundation models.

EAGLE-Net enhances foundation models by integrating patch-level features for better tissue understanding.

CDFM aims to unify causal discovery across diverse datasets.

We present a computational toolkit for (local) Poisson-Nijenhuis calculus on manifolds. Our python module $\textsf{PoissonGeometry}$ implements our algorithms, and accompanies this paper. We include two examples of how our methods can be used, one for gauge transformations of Poisson bivectors in dimension 3, and a sec…

GIFT-Eval benchmarks time series forecasting models across diverse datasets.

Unified approach to learn interpretable concepts from data.