New paper: Is the information geometry of probabilistic population codes learnable?

Manifold discovery in neural populations is hard. In our new paper, Is the information geometry of probabilistic population codes learnable?, we show that in a certain class of population activity models for which the latent manifold geometry can be inferred from the statistics of neural population activity. The work will be presented at the NeurReps workshop at NeurIPS. Check out John’s Twitter thread for more details.