Modified-MNIST

machine learning kaggle competition: https://www.kaggle.com/c/comp551-modified-mnist We have put our code all in jupyter notebook files and ran them to save you the time from running it.

The original set-up was to put the 3 given data sets(train_x.csv, train_y.csv, and test_x.csv) in the input folder we created, and you can execute any notebook/.py files without a problem doing so.

All our code are in the module folder. All our predictions are in csv inside the output folder.

Development Status:

• A baseline linear learner of SVM implemented using sklearn library: Done
• A fully connected feed forward neural network trained by backpropagation, where the architecture of the network (number of nodes, layers, learning rate, etc.) are determined by cross-validation: Done
• CNN implemented using pytorch library: Done

Members(Github names):

• DerekYu177
• jasondias9
• simonizerlol