Is there a way to use a range loop and record data for said loop for a scatter plot in python?

I'm making a basic Perceptron in Python 3.11.1 and I need a way to visualize change between epochs.

Here's my code:

import numpy as np
from matplotlib import pyplot as plt

def sigmoid(x):
    return 1 / (1 + np.exp(-x))

def sigmoid_d(x):
    return x * (1 - x)

training_inputs = np.array([[0, 0, 1],
                           [1, 1, 1],
                           [1, 0, 1],
                           [0, 1, 1]])

training_outputs = np.array([[1, 0, 1, 0]]).T

np.random.seed(1)

synaptic_weights = 2 * np.random.random((3, 1)) - 1

print("Random Starting Synaptic Weights: ")
print(synaptic_weights)

for i in range(100000):
    input_layer = training_inputs

    outputs = sigmoid(np.dot(input_layer, synaptic_weights))

    error = training_outputs - outputs

    adjustments = error * sigmoid_d(outputs)
    
    synaptic_weights += np.dot(input_layer.T, adjustments)
    
plt.scatter(i, outputs[0], c="blue")
plt.scatter(i, outputs[1], c="orange")
plt.scatter(i, outputs[2], c="green")
plt.scatter(i, outputs[3], c="red")

print("Synaptic Weights after training: ")
print(synaptic_weights)

print("Post-Training Outputs: ")
print(outputs)

plt.show()

When I tried it, it gave me the data from the last epoch. I was expecting for it to show the Perceptron training off of my basic dataset over time.

Solution

Change the indentation of the plt commands so that they are inside the loop and add a pause at each step to create an animation:

import numpy as np
from matplotlib import pyplot as plt

def sigmoid(x):
    return 1 / (1 + np.exp(-x))

def sigmoid_d(x):
    return x * (1 - x)

training_inputs = np.array([[0, 0, 1],
                           [1, 1, 1],
                           [1, 0, 1],
                           [0, 1, 1]])

training_outputs = np.array([[1, 0, 1, 0]]).T

np.random.seed(1)

synaptic_weights = 2 * np.random.random((3, 1)) - 1

print("Random Starting Synaptic Weights: ")
print(synaptic_weights)

for i in range(100):
    input_layer = training_inputs

    outputs = sigmoid(np.dot(input_layer, synaptic_weights))

    error = training_outputs - outputs

    adjustments = error * sigmoid_d(outputs)
    
    synaptic_weights += np.dot(input_layer.T, adjustments)
    
    plt.scatter(i, outputs[0], c="blue")
    plt.scatter(i, outputs[1], c="orange")
    plt.scatter(i, outputs[2], c="green")
    plt.scatter(i, outputs[3], c="red")
    plt.pause(0.05)

print("Synaptic Weights after training: ")
print(synaptic_weights)

print("Post-Training Outputs: ")
print(outputs)

plt.show()

The plot will not be updated "live" but at the end of the loop.

For real-time animation you could look at the Matplotlib animation API.