test.py

from MLP import MLPNetwork
import numpy as np
from time import perf_counter

# Use to get a slice of our data set


def sample(i, idx):
    return i[idx, :]


# Create our truth table dataset
input = np.zeros((16, 4))
output = np.zeros((16, 1))
for a in range(2):
    for b in range(2):
        for c in range(2):
            for d in range(2):
                input[8 * a + 4 * b + 2 * c + d, :] = np.array([a, b, c, d])
                output[8 * a + 4 * b + 2 * c + d, :] = (
                    1 if (a + b + c + d) % 2 == 1 else 0
                )
print(*[np.append(sample(input, i), output[i]) for i in range(16)], sep="\n")
# Validate our set
for i in range(16):
    row = sample(input, i)
    assert (1 if np.sum(row[:4]) % 2 == 1 else 0) == output[i], "Incorrect set"
print("All cases correct!")


def time(
    func, *params
):  # helper function to print the time it takes to execute a function
    start = perf_counter()
    result = func(*params)
    end = perf_counter()
    print(f"{func.__name__} took {end-start} seconds to complete.")
    return result


layers = [4, 4, 1]
nn = MLPNetwork(layers)
nn.display()

nn.train(input, output, 1000)

results = nn.predict(input[0, :])[0]
print(f"Model predicts: {np.round(results)}. (Expected: {output[0][0]}, got {results})")
nn.display()