How many classifiers are there in M3SDA and M3SDA-beta?

[yufengyin]

In the original M3SDA paper, page 4 (descriptions of M3SDA and M3SDA-beta), the paper says: "a set of N classifiers C = {C1, C2, ..., CN }" (M3SDA) and "we leverage two classifiers per domain to form N pairs of classifiers C' ={(C1, C1'),(C2, C2'), ...,(CN , CN')}."(MSDA-beta). It means that there should be N and 2N classifiers for M3SDA and M3SDA-beta.

However, in code_MSDA_digit/solver_MSDA.py:
Function train() shows that M3SDA has only 1 classifier. Function train_msda() shows that M3SDA-beta has only 2.

[seulchanson]

two, you should watch MCD_DA "Maximum Classifier Discrepancy for Unsupervised Domain Adaptation".