Center for Advanced Intelligence Project

This talk will be held in a hybrid format, both in person at AIP Open Space of RIKEN AIP (Nihonbashi office) and online by Zoom. AIP Open Space: *only available to AIP researchers.

DATE & TIME
Feb 18, 2025: 15:00 pm – 16:00 pm (JST)

TITLE
Machine Learning Models of Abstraction

SPEAKER
Prof. John Lafferty (Yale University)

ABSTRACT
Reasoning in terms of relations, analogies, and abstraction is a hallmark of human intelligence. This ability is largely separate from function approximation for sensory tasks such as image and audio processing. How can abstract symbols emerge from distributed, neural representations? One general approach uses an inductive bias for learning called the “relational bottleneck” that is motivated from principles of cognitive neuroscience. We present a framework that casts this inductive bias in terms of an extension of transformers, in which new types of attention mechanisms or kernels transform distributed symbols to implement a form of abstraction. Universal approximation theory is used to assess the flexibility of the approach, and complexity theory is used to study its limitations. Biologically plausible mechanisms are discussed to shed light on how models of abstraction could be implemented in the human brain.

BIOGRAPHY
John Lafferty is John C. Malone Professor in the Department of Statistics and Data Science at Yale University, with a secondary appointment in Computer Science. Lafferty is an Associate Director of the Wu Tsai Institute at Yale, a University-wide institute focused on the mission of understanding human cognition and exploring human potential by sparking interdisciplinary inquiry. He is Director of the Center for Neurocomputation and Machine Intelligence within the WTI. Lafferty’s most recent research is driven by the goal of using computational modeling, in particular machine learning, to gain insight into the remarkable abilities of the human brain. This computational lens can complement the specialized, biologically-grounded studies of traditional experimental science and mechanistic computational models. Lafferty’s research lab develops machine learning methodology together with theory that can help explain the behavior of the underlying algorithms.