2nd place solution for the 2017 national datascience bowl

Summary

This document describes my part of the 2nd prize solution to the Data Science Bowl 2017 hosted by Kaggle.com. I teamed up with Daniel Hammack. His part of the solution is decribed here The goal of the challenge was to predict the development of lung cancer in a patient given a set of CT images. Detailed descriptions of the challenge can be found on the Kaggle competition page and this blog post by Elias Vansteenkiste. My solution (and that of Daniel) was mainly based on nodule detectors with a 3D convolutional neural network architecture. I worked on a windows 64 system using the Keras library in combination with the just released windows version of TensorFlow.

3rd place solution for the second national datascience bowl

Summary

This document describes the 3rd prize solution to the Second National Data Science Bowl hosted by Kaggle.com. The goal of the challenge was to perform automatic volume measurement of the left ventricle based on MRI images. I will skip the description of the challenge since it has already been described on the Kaggle competition page and this blog post of the #2. My solution relied heavily on image segmentation with a neural network architecture called U-nets.

Click-through rate prediction using an ensemble of neural networks

Summary

This document describes the 4th prize solution to the Criteo Labs display advertising challenge hosted by Kaggle.com. The solution consisted of two steps. First the data was preprocessed. Rare and unseen test-set categorical values were all encoded as one category. The remaining features were one-hot encoded or hashed so that we ended up with roughly 200K separate sparse features. The second step was to use neural networks to train a number of different models with variations coming from different network architectures, bagging and preprocessing parameters. The different models were averaged into one final solution. test