构建两层神经网络分类器

This repository is the implementation of a handcrafted neural network model for classification of MNIST data.

Files in the folder

• mnist_data/
  • t10k-images-idx3-ubyte.gz
  • t10k-labels-idx1-ubyte.gz
  • train-images-idx3-ubyte.gz
  • train-labels-idx1-ubyte.gz
• models/
  • model_lr=0.001_hiddens=256_lambda2=1e-05.pkl
  • ...
• plots/
  • acc_Curve_lr=0.001_hiddens=256_lambda2=1e-05.jpg
  • Loss_Curve_lr=0.001_hiddens=256_lambda2=1e-05.jpg
  • weights_lr=0.001_hiddens=256_lambda2=1e-05.jpg
• main.py
• metrics.py
• neural_network.py
• preprocess.py

Required packages

The code has been tested running under Python 3.7.0, with the following packages installed (along with their dependencies):

• numpy == 1.15.1
• matplotlib == 2.2.3

Running Procedure

train the model

$ python main.py --mode=train --lr=0.001 --hiddens=256 --lambda2=0.00001

test the model

$ python main.py --mode=test --lr=0.001 --hiddens=256 --lambda2=0.00001

only if a model with the same hyperparameters are trained, it can be tested

the trained model will be saved in models/ and the plots will be saved in plots/