Mert Pilancı

Minimax Optimal Quantization of Linear Models: Information-Theoretic Limits and Efficient Algorithms

2022 · arXiv (Cornell University)

High-dimensional models often have a large memory footprint and must be quantized after training before being deployed on resource-constrained edge devices for inference tasks. In this work, we develop an information-theoretic framework for the problem …

Orthonormal Sketches for Secure Coded Regression

2022 · arXiv (Cornell University)

In this work, we propose a method for speeding up linear regression distributively, while ensuring security. We leverage randomized sketching techniques, and improve straggler resilience in asynchronous systems. Specifically, we apply a random orthonormal matrix …

Active Learning of Deep Neural Networks via Gradient-Free Cutting Planes

2024 · arXiv (Cornell University)

Active learning methods aim to improve sample complexity in machine learning. In this work, we investigate an active learning scheme via a novel gradient-free cutting-plane training method for ReLU networks of arbitrary depth and develop …

LLM-Lasso: A Robust Framework for Domain-Informed Feature Selection and Regularization

2025 · arXiv (Cornell University)

We introduce LLM-Lasso, a novel framework that leverages large language models (LLMs) to guide feature selection in Lasso $\ell_1$ regression. Unlike traditional methods that rely solely on numerical data, LLM-Lasso incorporates domain-specific knowledge extracted from …