Think about distance metric vs comparison metric for RDM

[CeliaBenquet]

here:

CEBRA-lens/cebra_lens/quantification/rdm_metric.py

Line 25 in fce191a

metric : str, optional

Currently, correlation is implemented (the only option) for

• computing the RDM.
• and comparing multiple RDMs.

Computing the RDM
do we want more distances? pdist handles:

metricstr or function, optional
The distance metric to use. The distance function can be ‘braycurtis’, ‘canberra’, ‘chebyshev’, ‘cityblock’, 
‘correlation’, ‘cosine’, ‘dice’, ‘euclidean’, ‘hamming’, ‘jaccard’, ‘jensenshannon’, ‘kulczynski1’, ‘mahalanobis’, 
‘matching’, ‘minkowski’, ‘rogerstanimoto’, ‘russellrao’, ‘seuclidean’, ‘sokalmichener’, ‘sokalsneath’, ‘sqeuclidean’, ‘yule’.

in https://rsatoolbox.readthedocs.io/en/stable/distances.html#correlation-distance

The correlation distance is part of RSA since the start and was found to be similarly reliable as the Euclidean-like dissimilarities (Walther et. al 2016). The interpretation of correlation distances is harder than for the euclidean types though. The Euclidean-like dissimilarities all depend only on the difference between the two patterns. In contrast, the correlation additionally depends on shared overall activations, such that an additional shared activity which does not impair decoding still reduces the correlation distance.

Comparing multiple RDMs
and do we want a different measure of similarity: https://rsatoolbox.readthedocs.io/en/stable/comparing.html.