A new flow matching framework unifies various distance metrics as Jeffreys divergences under different velocity constraints. This approach improves the estimation of distances between complex neural response distributions and continuous variables.
HOW THIS AFFECTS YOU
●
researcherYou can use flow matching to design and estimate more principled neural distance metrics.