This repository demonstrates the implementation of various deep learning models using PyTorch, focusing on regression and classification tasks.

This project implements three main types of models:

• Linear Regression
• Binary Classification (Moon Dataset)
• Multi-class Classification (Spiral Dataset)

A simple linear regression model implementing y = Wx + b

• Loss Function: L1Loss (Mean Absolute Error)

  L1(y, ŷ) = |y - ŷ|
  

• Optimizer: Stochastic Gradient Descent (SGD)

  w = w - learning_rate * gradient
  

• Learning Rate: 0.01
• Epochs: 300

Implementation of binary classification using a neural network on the make_moons dataset.

Sequential(
    Linear(2 → 10)
    ReLU()
    Linear(10 → 10)
    ReLU()
    Linear(10 → 10)
    ReLU()
    Linear(10 → 1)
)

• Loss Function: Binary Cross Entropy with Logits

  BCE(x, y) = -[y * log(σ(x)) + (1 - y) * log(1 - σ(x))]
  

• Optimizer: SGD
• Learning Rate: 0.1
• Epochs: 1000

Implementation of multi-class classification on a spiral dataset.

Sequential(
    Linear(2 → 10)
    ReLU()
    Linear(10 → 10)
    ReLU()
    Linear(10 → 10)
    ReLU()
    Linear(10 → 3)
)

• Loss Function: Cross Entropy Loss

  CE(x, y) = -Σ y_i * log(softmax(x_i))
  

• Optimizer: Adam

  m_t = β_1 * m_{t-1} + (1 - β_1) * g_t
  v_t = β_2 * v_{t-1} + (1 - β_2) * g_t^2
  

• Learning Rate: 0.1
• Epochs: 200

• Python 3.8+
• PyTorch
• scikit-learn
• matplotlib
• numpy

git clone https://github.com/yourusername/pytorch-regression-classification.git
cd pytorch-regression-classification
pip install -r requirements.txt

jupyter notebook notebook.ipynb

• Implementation of three different types of neural networks
• Visualization of decision boundaries
• Loss curve tracking and visualization
• Model performance analysis
• Comprehensive documentation

• PyTorch - Deep Learning Framework
• scikit-learn - Dataset Generation
• Matplotlib - Visualization
• NumPy - Numerical Computations

This project is licensed under the MIT License - see the LICENSE.md file for details

Contributions, issues, and feature requests are welcome! Feel free to check issues page.