deep_learning_from_scratch_exercise_2.py

deep_learning_from_scratch_exercise_2.py

import pandas as pd
import torch

"""
COMPLETE THE CODE BELOW BY REPLACING THE QUESTION MARKS ?'s
SO FORWARD PROPAGATION IS COMPLETE 
"""


url = "https://raw.githubusercontent.com/thomasnield/machine-learning-demo-data/master/classification/maintenance_predict.csv"
df = pd.read_csv(url)

# Extract the input columns, scale down by 255, and convert to PyTorch tensors
# Using float() ensures compatibility with weights during matrix multiplication
X = torch.tensor(df.iloc[:, 0:3].values, dtype=torch.float32) / 255.0
Y = torch.tensor(df.iloc[:, -1].values, dtype=torch.float32)

# Build neural network with weights and biases
# with random initialization
w_hidden = torch.rand(3, 3)
w_output = torch.rand(1, 3)

b_hidden = torch.rand(3, 1)
b_output = torch.rand(1, 1)

# Activation functions
relu = lambda x: torch.relu(x)
logistic = lambda x: torch.sigmoid(x)


# Runs inputs through the neural network to get predicted outputs
def forward_prop(X):
    Z1 = ? @ X + b_hidden
    A1 = relu(?)
    Z2 = ? @ ? + b_output
    A2 = logistic(?)
    return Z1, A1, Z2, A2


# Calculate accuracy
# Note: X.t() is the shorthand transpose method for 2D tensors in PyTorch
test_predictions = forward_prop(X.t())[3]

# .flatten() works similarly; .to(torch.int) replaces .astype(int)
test_comparisons = torch.eq(
    (test_predictions >= 0.5).flatten().to(torch.int), Y.to(torch.int)
)
accuracy = torch.sum(test_comparisons.to(torch.float32)) / X.shape[0]

print("ACCURACY: ", accuracy.item())