Model Performance Evaluation via Contrastive Learning of Distilled Surrogate Models

Makoto Kawano, Kazuki Kawamura

Abstract

This work evaluates model performance through contrastive learning with distilled surrogate models.

Type
Conference paper
Publication
The 38th Annual Conference of the Japanese Society for Artificial Intelligence (JSAI2024)
Representation Learning Model Evaluation
Makoto Kawano
Project Assistant Professor, Matsuo Laboratory, The University of Tokyo

Researcher working on transfer learning, deep generative models, foundation models, robotics, autonomous driving, and machine learning for real-world data.