Author : Yuguang Lee -- University of Washington
Email : [email protected]
This project is a part of my machine learning class from the University of Washington. The goal is to implement different machine learning algorithms to classify the handwritten data. The project includes four chapters including multiple parts as chapter name. To run the program, you'll need to download the MNIST handwritten datasets from the link : http://yann.lecun.com/exdb/mnist/ . Python codes are included in the toolbox to load the dataset.
The best performance (Accuracy 99%) of classiying 10 digits is found in SVM with linear kernels. But it runs fairly slow. The neural nets (with relu kernel) trains in 3min and gives us an accuracy of 97.9%.