Contributed Talk Sessions | Poster Sessions | All Posters | Search Papers

Poster Session A: Tuesday, August 12, 1:30 – 4:30 pm, de Brug & E‑Hall

Framed RSA: honoring representational geometry and regional-mean response preferences

JohnMark Taylor¹, Nikolaus Kriegeskorte¹; ¹Columbia University

Presenter: JohnMark Taylor

Representational similarity analysis (RSA) characterizes the geometry of neural activity patterns elicited by different stimuli while discarding their regional-mean activity and the location or orientation of the patterns in multivariate response space. Regional-mean activation analysis serves the complementary purpose of comparing the average population response to different stimuli. Here we introduce a novel method, framed RSA, which honors both the geometry and the regional-mean preferences in evaluating model-predicted representations. To achieve this, we augment the stimulus patterns with two reference patterns: the zero-point (origin) and a uniform constant pattern, enabling RSA to incorporate information about the global location, orientation, and mean activation of neural population codes. Framed RSA improves accuracy for both brain region identification (using fMRI data from the Natural Scenes Dataset) and deep neural network layer identification relative to existing RSA approaches. Framed RSA thus combines the strengths of two complementary and traditionally separate analysis approaches, and improves power for model-comparative inference.

Topic Area: Methods & Computational Tools

Extended Abstract: Full Text PDF