Topic: Machine Learning Neural Networks

In this homework, you will be required to implement Fully Connected Neural Network (FC-NN) to recognize the handwriting numbers (0-9). We will use MNIST database as our training data set and test data set. The information of the database can be found through http://yann.lecun.com/exdb/mnist/. Please look at the website carefully.

You are required to use FC-NN to learn the data. There are three layers in this FC-NN: input layer, output layer, and a single hidden layer. You should determine how many nodes at each layer.

Use either Java or Python. Though I think Python would be a lot easier for this.

Questions:

1. how many nodes are there at each layer

2. how well of your learning output in terms of the error rate rate

3. what is the good/clever part you think in your implementation

4. what is the difficult part in your implementation