Visual feature manifolds in a convolutional RNN

We compared manifold capacity and other related measures (see: https://www.nature.com/articles/s41467-020-14578-5) with representational similarity measures from the RSA toolbox (https://www.frontiersin.org/journals/systems-neuroscience/articles/10.3389/neuro.06.004.2008/full) for dynamic representations from a recurrent convolutional neural network performing object recognition, and found interesting divergences that remain to be studied in detail.

Work done at the 2023 Analytical Connectionism Workshop at UCL.