Progress on the 2021 Generative Adversarial Collaboration "What constitutes understanding of ventral pathway function?"

Carlos Ponce, Gabriel Kreiman, Harvard Medical School, United States

GAC Posters: Progress updates Poster

Late-breaking Research

Pacific Ballroom H-O

Presentation Time:
Thu, 25 Aug, 19:30 - 21:30 Pacific Time (UTC -8)

An exciting challenge in cognitive computational neuroscience is to learn how the primate brain realizes visual recognition. In 2021, a group of vision scientists organized a Generative Adversarial Collaboration to identify the most promising ways to study the visual recognition network, known as the ventral stream. Although the workshop outlined many useful and validated approaches, none of them were perceived as able to address all major questions. In follow-up meetings, a core set of teams worked to craft a series of experimental exercises (or challenges), each designed to observe basic criteria of falsifiability, generalization, and interpretability. Here, we describe these challenges and overall repeating themes that emerged during the ongoing collaboration. Each theme is characterized as a heuristic — a practical method not guaranteed to be optimal, but still sufficient for achieving answers. We break down each heuristic into an (1) objective methodology and (2) a more subjective explanatory incentive. We also illustrate how the challenges tie together some of the major heuristics, as well as the many years of future investigation that this collaboration has borne out.
No resources available.