Congratulations on your choice of a interesting course if you find this repository. Don't be tired of his f**g lectures, or you would be failed in the final exam...

• features: output/data/dataset_train.npy dataset_test.npy (3558 * num_of_features)
• labels: output/data/target_train.npy dataset_test.npy (3558 * 1)

using 1 fc-layer.

using 1 fc-layer and hinge loss.

using several fc-layers and crossentropy loss.

• LR: 88% (with l2 norm) 87% (without norm)
• SVM: 89%
• NN: 91% (with l2 norm, dropout(0.5), relu activation funciton)

Results could be improved with some more parameters adjustment.