Michael W. Mahoney

End-to-End Probabilistic Framework for Learning with Hard Constraints

Understanding Transformers for Time Series: Rank Structure, Flow-of-ranks, and Compressibility

Uncertainty-Aware Diagnostics for Physics-Informed Machine Learning

The False Promise of Zero-Shot Super-Resolution in Machine-Learned Operators

Understanding the Implicit Biases of Design Choices for Time Series Foundation Models

Zero-shot Forecasting by Simulation Alone

Time-Aware Prior Fitted Networks for Zero-Shot Forecasting with Exogenous Variables

When Does Multimodality Lead to Better Time Series Forecasting?

Fundamental Bias in Inverting Random Sampling Matrices with Application to Sub-sampled Newton

Tuning Frequency Bias of State Space Models

Mitigating Memorization in Language Models

Spectral Estimation with Free Decompression

Mitra: Mixed Synthetic Priors for Enhancing Tabular Foundation Models

Determinant Estimation under Memory Constraints and Neural Scaling Laws

Multipole Attention for Efficient Long Context Reasoning

HOPE for a Robust Parameterization of Long-memory State Space Models

Enhancing Foundation Models for Time Series Forecasting via Wavelet-based Tokenization

QuantSpec: Self-Speculative Decoding with Hierarchical Quantized KV Cache

A Statistical Framework for Ranking LLM-based Chatbots

Models of Heavy-Tailed Mechanistic Universality

Enhancing Foundation Models for Time Series Forecasting via Wavelet-based Tokenization

Gradient-Free Generation for Hard-Constrained Systems

Elucidating the Design Choice of Probability Paths in Flow Matching for Forecasting

Comparing and Contrasting Deep Learning Weather Prediction Backbones on Navier-Stokes and Atmospheric Dynamics