This repo contains some jupyter notebooks used in kaggle competitions
-
digits: use convolutional neural network, data augmentation to recogize hand written digit, achieved higher than 98% accuracy.
-
houseprice: explored linear regression, lasso regression, ridge regression, elastic net regression in predicting house price. Tackled with feature engineering problems such as missing data, skewed data, new feature synthesize, important feature selection.
-
titanic: explored random forest, support vector machine, gradient boosting, k nearest neighbor in predicting whether a passenger survived in titanic.
-
nuclei: use U-net (see: Ronneberger, O.; Fischer, P.; Brox, T., 2015. U-Net: Convolutional Networks for Biomedical Image Segmentation) to do semantic segmentation on biomedical images to recognize cell nuclei.