This repo contains the main functionality from my entry in the Kaggle competition, Otto Group Product Classification Challenge.

The code links togehter the neural network library Lasagne with the hyperparameter optimization package Hyperopt. The code in utils provide everything needed to build and train a neural network. The script sqrttrans_pca.py runs hyperopt in order to optimize hyperparameters.

##Dependencies

• Lasagne
• Numpy
• Hyperopt
• Sklearn

Note: it is probably not easy to get this running, because the code is from May 2015 and Lasagne is under rapid development. In addition hyperopt needs some special attention as it needs an old version of pymongo.

Also the data need to be downloaded from Kaggle.