IzPerfect/MNIST-MLP

Implementation of MLP in Tensorflow

Multilayer Perceptron(MLP)

Tensorflow implementaion of MLP for classification of MNIST_dataset. Basic code for studying.

Requirement

1. Tensorflow 1.4.0
2. Python 3.5.4
3. Python packages : numpy, matplotlib, os, argparse

Usage

Command

python main.py

Arguments

Optional

• --layer1 : Number of layer1 nodes in MLP. Default : 64
• --layer2 : Number of layer2 nodes in MLP. Default : 128
• --layer3 : Number of layer3 nodes in MLP. Default : 256
• --epoch : Number of epochs to run. Default : 10
• --batch_size : Number of batch_size to run. Default : 50
• --learning_rate : Learning rate for Adam optimizer. Default : 0.001
• --drop_rate : Prob of dropout. Default : 0.7
• --disp_num : How many display MNIST prediction. Default : 5

Results

python main.py

training

Prediction number and test image

Reference Implementations

• https://github.com/golbin/TensorFlow-ML-Exercises