Date: June 20, 2025

Speaker: Dr. Peer Herholzt

Topic: Reproducible deep and machine learning

Location: Conference Room 503, Teaching and Research Building, Shuang-Ho Campus

In addition to the three-day Open Mind workshop, the Institute of Mind, Brain and Consciousness organized by Director Dr. Tzu-Yu Hsu and Associate Professor Niall Duncan once again invited Dr. Peer Herholz (Assistant Professor at Northwestern University, USA) to give a talk to faculty and students.

Dr. Herholz’s research focuses on two main areas: first, using auditory processing as a model to explore the neurobiology of complex biological behavior; and second, combining computational and conceptual models to study behavioral generalization and the underlying transformation mechanisms in both biological and artificial intelligence systems.

The content addressed a fundamental issue present in most machine learning research: the lack of reproducibility. Even when the same model is run twice on the same computer, the results may differ. Dr. Herholz introduced tools and methods aimed at addressing this problem, while also elaborating on the conceptual foundations that make true reproducibility in machine learning and deep learning nearly impossible.