Instructions:
• Adjust the $\alpha$ slider to change the slope
• Adjust the $\beta$ slider to move the decision boundary (with sigmoid) or intercept (without sigmoid)
• Toggle sigmoid activation to see the effect on predictions
• Click above y=0.5 to add class 1 points, below y=0.5 to add class 0 points
• Use "Generate New Data" to create random binary classification data
• Use "Clear All Data" to remove all points
• Find Optimal Solution: Without sigmoid: minimizes MSE using least squares. With sigmoid: minimizes cross-entropy using gradient descent

Metrics:
• Accuracy: Percentage of correctly classified points (same for both modes)
• Loss: Without sigmoid: Mean Squared Error (MSE) between predictions and true labels
• Loss: With sigmoid: Cross-entropy loss, the standard loss function for classification