test.py

import random
import numpy as np
import matplotlib.pyplot as plt
from nn import *  # Import your neural network functions
from data import mnist_dataloader

# Load the trained parameters

packed_params = np.load('trained_params_with_mini_batch8.npy')

dimensions = [784,100,10]

W1, b1, W2, b2 = unpack_params(packed_params, dimensions)

def predict(X):
    # Forward pass
    Z1 = W1 @ X + b1
    A1 = sigmoid(Z1)
    Z2 = W2 @ A1 + b2
    Y_hat = softmax(Z2)
    predicted_class = np.argmax(Y_hat, axis=0)
    return predicted_class

# Normalize and preprocess data (ensure x_test is normalized)
(x_train, y_train), (x_test, y_test) = mnist_dataloader.load_data()
x_test = np.array(x_test) / 255.0
x_test = x_test.reshape(x_test.shape[0], -1).T  # Reshape for prediction
y_test = np.array(y_test)
# If y_test is one-hot encoded, convert to class labels
if y_test.ndim > 1:  # Check if y_test is one-hot encoded
    y_test = np.argmax(y_test, axis=0)

# Make predictions on the entire test set
predicted_classes = predict(x_test)

# Calculate accuracy
correct_predictions = np.sum(predicted_classes == y_test)
accuracy = correct_predictions / x_test.shape[1] * 100  # x_test.shape[1] is the number of test samples
print(f"Test Accuracy: {accuracy:.2f}%")

# Optionally, display a random test image and its prediction
random_index = random.randint(0, x_test.shape[1] - 1)  # Random test index
sample_image = x_test[:, random_index].reshape(28, 28)  # Reshape to 28x28 for visualization
for i in range(28):
    for j in range(28):
        if sample_image[i][j] == 0:
            print(" ", end=" ")  # Replace zero with a space
        else:
            print("0", end=" ")  # Replace non-zero with 0
    print()  # Moves to the next line after printing a row
plt.imshow(sample_image, cmap='gray')
plt.title(f"Sample Index: {random_index}, Predicted: {predicted_classes[random_index]}")
plt.show()