2.5 Python implementation

Python implementation

🎙️Marcelo Gutiérrez

Description

This section provides a brief introduction to the perceptron algorithm and the AND logic gate dataset that we will be using in this tutorial.

Network Initialization

import pandas as pd
import numpy as np

data = [[-1,-1, 1,-1],
        [-1, 1, 1,-1],
        [ 1,-1, 1,-1],
        [ 1, 1, 1, 1]]

pd.DataFrame(data, index=list(range(len(data))), columns=['X1', 'X2','bias','Y'])

Output

### See secction "2.2 Perceptron evolution"###
prediction = sign(np.dot(weight, training))

Activation function

### See secction "2.2 Perceptron evolution"###
def sign(x):
    if x > 0:
        return 1.0
    elif x < 0:
        return -1.0
    else:
        return 1.0

Error of the neural network model

### Ver sección "2.4 Loss function"###
error = desired - prediction
sse += error**2

Weight update

### Ver sección "2.3 Teach To Learn"###
weight = weight + error * training

Algorithm

def sign(x):
    if x > 0:
        return 1.0
    elif x < 0:
        return -1.0
    else:
        return 1.0
    
import pandas as pd
import numpy as np

data = [[-1,-1, 1,-1],
        [-1, 1, 1,-1],
        [ 1,-1, 1,-1],
        [ 1, 1, 1, 1]]

pd.DataFrame(data, index=list(range(len(data))), columns=['X1', 'X2','bias','Y'])
train_data = pd.DataFrame(data, index=list(range(len(data))), columns=['X1', 'X2','bias','Y'])
train_Y = train_data.Y
train_data = train_data[['X1', 'X2','bias']]
train_data.head()

weight = np.zeros(train_data.shape[1])

for epoch in range(100):
    sse = 0.0
    for sample in range(train_data.shape[0]):
        training = train_data.loc[sample].values
        desired = train_Y.loc[sample]
        prediction = sign(np.dot(weight, training))
        error = desired - prediction
        sse += error**2
        weight = weight + error * training
    print(f'SSE: {sse}')
    if sse == 0:
        break
print(f'Weights: {weight}, Epoch: {epoch+1}')