Random Forest baseline

[dkoutsou]

Random Forest baseline

[melanibe]

For information count per label on the training set:

array([12885, 1254, 3621, 1561, 1858, 2513, 1008, 2822, 53,
45, 28, 1093, 688, 537, 1066, 21, 530, 210,
902, 1482, 172, 3777, 802, 2965, 322, 8228, 328,
11])

[Minimel]

Tried data augmentation, with number of times I rotate and reverse image depending on target label (more if target label rarer) to get more even class distribution. Tried to parallelize on Euler because takes some time. Will see how this affects baseline.

[ag14774]

I'll try a few things for this one today or tomorrow. I'll try oversampling too