Extended Data Fig. 3: Embedding-invariant representations of convergent and divergent vector fields.
From: MARBLE: interpretable representations of neural population dynamics using geometric deep learning
a Convergent and divergent vector fields sampled uniformly at random (n=512) in the interval [−1, 1]2. b Embedding-agnostic MARBLE representations can distinguish the fields, even without rotational information. Black lines show k-means clustering (k=15). c Histogram of LFF types confirming the disjoint representations of the two flow fields. d One LFF drawn randomly from each cluster displays expanding LFFs for types 1–8 and contracting LFFs for types 9-15.
