Dmitrii Tsybulevskii

I've also had some problems with this model: it gaves me wrong predictions with cudnn v3. mxnet without cudnn, with cudnn v4, and cpu version worked fine for me.

Sorry for the late reply. Thanks for your interest. In general you are right, but I've used 7 different feature sets, described [here](blog.kaggle.com/2016/04/28/yelp-restaurant-photo-classification-winners-interview-1st-place-dmitrii-tsybulevskii/#how-did-you-deal-with-the-multi-instance-aspect-of-this-problem), and for the some of feature sets...

Hi Pawan, I've used slightly modified version Classifier Chain from [scikit-learn PR](https://github.com/scikit-learn/scikit-learn/pull/3727). Some kind of problems may be due to the type of predictions: 0,1 labels, or probability. Could you...