VLAI-AIVN/AIO2025M06_DEMO_LOGISTIC_REGRESSION

Logistic Regression Demo

Interactive demonstration of Logistic Regression implemented from scratch using NumPy and gradient descent. Learn binary classification with sigmoid activation, binary cross-entropy loss, and adjustable prediction threshold.

Features

• —Binary Classification: Implements binary classification (2 classes: 0 and 1)
• —NumPy Implementation: Efficient matrix operations for fast computation
• —Sigmoid Activation: Maps predictions to probabilities (0-1 range)
• —Binary Cross-Entropy Loss: Optimized loss function for binary classification
• —Adjustable Threshold: Experiment with different probability thresholds to balance precision/recall
• —Mini-batch Gradient Descent: Supports configurable batch sizes (powers of 2) or full batch
• —Feature Normalization: Automatic standardization (zero mean, unit variance) for stable training
• —Training Visualization: Track loss and accuracy over epochs for training and validation sets

Algorithm Details

Activation Function: Sigmoid σ(z) = 1/(1 + e^(-z)) Loss Function: Binary Cross-Entropy L = -[y·log(ŷ) + (1-y)·log(1-ŷ)] Classification: Predict class 1 if probability ≥ threshold, else class 0 Normalization: Features standardized (zero mean, unit variance) for numerical stability

Sample Datasets

1. 1.Breast Cancer: Wisconsin Breast Cancer dataset (binary classification)
2. 2.Wine (Binary): Wine dataset converted to binary (class 0 vs others)
3. 3.Synthetic: Artificially generated binary classification dataset

How to Use

1. 1.Select Data: Choose a sample dataset or upload your own CSV/Excel file
2. 2.Configure Target: Select target column (must have exactly 2 unique values)
3. 3.Set Training Parameters:
4. 4.Epochs: Number of training iterations (recommended: 50-500)
5. 5.Learning Rate: Step size for gradient descent (recommended: 0.001-0.01)
6. 6.Batch Size: Samples per batch (powers of 2, or Full Batch)
7. 7.Train/Validation Split: Proportion for training (default: 80%)
8. 8.Adjust Threshold: Set probability threshold for classification (default: 0.5)
9. 9.Enter Features: Input feature values for prediction
10. 10.Run Training: Click "Run Training & Prediction" to train and visualize

Key Parameters

Training Parameters:

• —Epochs: Complete passes through data. More epochs = better learning but risk of overfitting
• —Learning Rate: Step size (0.001-0.01 recommended). Too high causes instability, too low is slow
• —Batch Size: Samples processed before update. Smaller = faster but noisier, larger = more stable
• —Train/Validation Split: Data split ratio (default 80/20)

Threshold Parameter (Key Feature):

• —Default: 0.5 (balanced classification)
• —Lower threshold (e.g., 0.3): More class 1 predictions → higher recall, lower precision
• —Higher threshold (e.g., 0.7): Fewer class 1 predictions → higher precision, lower recall
• —Experiment: Adjust threshold to see how predictions and accuracy change in real-time
• —Use Case: Balance precision vs recall based on your classification goals

Requirements

• —gradio >= 5.38.0
• —pandas >= 1.5.0
• —scikit-learn >= 1.3.0
• —numpy >= 1.24.0
• —plotly >= 5.15.0