Project: AND Logic Gate using scikit-learn

In this demonstration, we train a Perceptron using scikit-learn to learn the AND logic gate. Unlike the manual implementation, scikit-learn automatically handles weight initialization, iterative training, and convergence.

Step 1: Define the Training Dataset

Create the inputs and outputs for the AND function.

Toggle Code


import numpy as np
from sklearn.linear_model import Perceptron

# Inputs (x1, x2) and targets (AND truth table)
X = np.array([
  [0, 0],
  [0, 1],
  [1, 0],
  [1, 1]
])
y = np.array([0, 0, 0, 1])  # AND gate outputs

Explanation:

Each row in X represents a possible combination of inputs.
y contains the target outputs corresponding to the AND gate.
This format satisfies scikit-learn’s (n_samples, n_features) requirement.

Step 2: Initialize and Train the Perceptron

Initialize a scikit-learn Perceptron and train it.

Toggle Code


# Initialize Perceptron
perceptron = Perceptron(
  max_iter=1000,    # Maximum training iterations
  eta0=0.1,         # Learning rate
  random_state=42)  # For reproducibility

# Train the Perceptron
perceptron.fit(X, y)

Explanation:

scikit-learn handles forward pass, error calculation, and weight updates internally.
Training continues until convergence or max_iter is reached.

Step 3: Inspect the Trained Model

Print and analyze the learned weights and bias.

Toggle Code


print("Perceptron coefficients (weights):", perceptron.coef_)
print("Perceptron intercept (bias):", perceptron.intercept_)

Explanation:

coef_ shows the importance of each input feature.
intercept_ is the bias that adjusts the decision threshold.

Step 4: Test the Trained Perceptron

Verify the output for all training inputs.

Toggle Code


for i in range(len(X)):
  y_pred = perceptron.predict(X[i].reshape(1, -1))[0]
  print(f"Input: {X[i]} → Output: {y_pred}")

Explanation:

predict() handles forward pass and thresholding internally.
Output is 1 only when both inputs are 1.

Step 5: Visualize the Decision Boundary

Plot the learned decision boundary.

Toggle Code


import matplotlib.pyplot as plt

# Plot points
plt.scatter(X[:, 0], X[:, 1], c=y, cmap=plt.cm.coolwarm, edgecolors='k')

# Decision boundary: w1*x1 + w2*x2 + b = 0
x_vals = np.array([X[:,0].min() - 0.5, X[:,0].max() + 0.5])
y_vals = -(perceptron.coef_[0][0] * x_vals + perceptron.intercept_[0]) / perceptron.coef_[0][1]

plt.plot(x_vals, y_vals, 'k--')
plt.xlabel("Input 1")
plt.ylabel("Input 2")
plt.title("Perceptron Decision Boundary (AND Gate)")
plt.show()

Explanation:

The line separates output 0 from 1.
All points are correctly classified on the appropriate side.

Understanding the Process

Data Setup: Each row is an AND truth table input.
Initialization & Training: scikit-learn initializes weights and updates them automatically.
Output: The model outputs 1 only for (1,1).
Visualization: Confirms the linear decision boundary separates the classes.

Outcome

The scikit-learn Perceptron converges quickly and correctly models the AND gate. This demonstrates that the AND gate is linearly separable and can be solved with a single-layer perceptron.