A Flow Matching Framework for Neural Representational Dissimilarity
Flow matching unifies neural representational distances as Jeffreys divergences under velocity constraints.
The paper unifies several neural representational dissimilarity metrics inside a flow-matching framework from deep generative models. It shows these distances arise as Jeffreys divergences that differ by the velocity constraints imposed on the flow. The authors argue the approach estimates distances more effectively for complicated distributions and continuous variables and gives a principled route to new metrics. The result is one framework for understanding, estimating, and designing comparisons of neural response distributions.
- Common distances arise as Jeffreys divergences under velocity constraints.
- Flow matching suits complex distributions and continuous variables.
- The same framework can design new representational distance metrics.
From the original article
1 min read at arxiv.orgNeural representational dissimilarity quantifies differences between neural response distributions, and is essential for comparing neural codes across stimuli, brain areas, tasks, and models. Commonly used distance metrics involve different assumptions and are estimated with separate methods.
By Zeyuan Ye · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.