色中色

Topic 

Towards Self-Improving Machine Learning

Abstract

Training data has become the critical driver of performance in machine learning. Yet, the process of constructing effective datasets remains largely manual, guided by heuristics, trial-and-error, and labor-intensive annotation. This challenge is especially pronounced in domains with limited data availability, such as complex reasoning tasks. In this talk, I will present the self-improving system we are developing at Fujitsu Research. Our system accelerates data discovery and automates the assembly and curation of training datasets, enabling continuous improvement in model performance with minimal human intervention.

About the speaker

Xavier Boix is the Director of Research of the Self-Improving Machine Learning group at Fujitsu Research of America. Prior to that, he was a research scientist at MIT, where he led a research group that investigated biologically-inspired machine learning. He earned his PhD in machine learning from ETH Zurich in 2014 and continued his postdoctoral training at MIT as part of the multidisciplinary Center for Brains, Minds, and Machines. Xavier鈥檚 research focuses on building the next generation of intelligent machines while also seeking to understand them. He integrates insights and techniques from theoretical machine learning, deep neural network engineering, and neuroscience to achieve this goal.

Your host

, Assistant Professor Data Mining at the department of Mathematics and Computer Science of the 色中色.

She will host Professor Xavier Boix, Director of Research of the Self-Improving Machine Learning group at Fujitsu Research of America.

Registration is required but free of charge.