hhkwon/NeuralNetwork

:triangular_ruler: NeuralNetwork with Python

NeuralNetwork with Python

run neuralnetwork to train mnist

What is this?

신경망을 직접 구현하고, mnist 데이터셋을 학습시킵니다.

• NeuralNetwork with Python : https://github.com/hhkwon/NeuralNetwork

Language

Python 3.6

DataSet

• Raw Data
  • mnist_test.csv ( mnist test set )
  • mnist_train.csv ( mnist train set )

Process

1. Sourcing Raw Data

2. Pre-processing

   • inputs = (np.asfarray(all_values[1:]) / 255.0 * 0.99) + 0.01

3. Modeling (Update Weight)

   • inputs = np.array(inputs_list, ndmin=2).T
   • targets = np.array(targets_list, ndmin=2).T
   • (hidden_inputs, hidden_outputs), (final_inputs, final_outputs)
   • output_errors = targets - final_outputs,
   • hidden_errors = np.dot(self.who.T, output_errors)
   • self.who += self.learning_rate * np.dot((output_errors * final_outputs * (1.0 - final_outputs)), np.transpose(hidden_outputs))
   • self.wih += self.learning_rate * np.dot((hidden_errors * hidden_outputs * (1.0 - hidden_outputs)), np.transpose(inputs))

4. Output Sharing

   • correct_label = int(all_values[0])
   • label = np.argmax(outputs)
   • scorecard_array = np.asarray(scorecard)
   • print("performance = ", scorecard_array.sum() / scorecard_array.size)

Library

`PYTHON`
import numpy as np
import matplotlib.pyplot as plt
import scipy.special

Release History

• 0.0.1
  • First Commit (2019/02/02)