[MRG] replace +1 by inf in MMC diag

[wdevazelhes]

This is a quick fix to a problem that was making fail the following example for diagonal MMC:

In [9]:  from sklearn.datasets import fetch_lfw_pairs 
   ...:  from sklearn.model_selection import cross_validate, train_test_split 
   ...:  from metric_learn import MMC 
   ...:  pairs, y_pairs = [fetch_lfw_pairs()[key] for key in ['pairs', 'target']] 
   ...:  pairs, _, y_pairs, _ = train_test_split(pairs, 2*y_pairs-1) 
   ...:  pairs = pairs.reshape(pairs.shape[0], 2, -1) 
   ...:  mmc = MMC(diagonal=True)
   ...:  mmc.fit(pairs, y_pairs)

what happens is that obj_previous (see here:

metric-learn/metric_learn/mmc.py

Line 207 in 899ef47

obj_previous = obj + 1 # just to get the while-loop started

) was such a big number (of order 10**20) that adding +1 made it stay the same number (see picture) and therefore the strict comparison failed to start the loop and therefore w_previous was unreferenced. Replacing obj_previous +1 by np.inf fixed the pb (it's a quick fix to make it work, but I don't really know if it's the final answer, what if obj_previous == np.inf too ? The comparison will then fail again. Should we detect this case and throw an error ?) I guess for the next release we can merge this as is though, and maybe investigate later ?

[bellet]

Thanks. The solution sounds good. Personally I don't think it is worth covering the case where the initial objective is infinity with a proper error, as this is very unlikely to happen. Indeed we already check that the data does not contain any inf value, and float64 can handle values as large as 1e300 or so.

[perimosocordiae]

Thanks, merged.