Skip to content
ELFAKIR-tahaPublic

About

Machine Learning TP exercises (Ridge, SVM, Classification, Kernels)

Resources

Stars

0 stars

Watchers

0 watching

Forks

Repository files navigation

Machine Learning - Practical Labs

Practical assignments for Machine Learning course at Télécom Saint-Étienne

Overview

This repository contains 4 hands-on machine learning labs covering fundamental concepts:

TP1 - Hyperparameter Selection

Ridge and Lasso regression with hyperparameter tuning on superconductor critical temperature dataset.

  • Dataset: data_DRISI.csv (superconductor properties)
  • Skills: Regularization, grid search, SGD training
  • Key Methods: Ridge regression, Lasso, cross-validation

TP2 - Multiclass Classification

Logistic regression for 3-class penguin species classification (Adélie, Chinstrap, Gentoo).

  • Dataset: penguins.csv (Palmer Penguins, 2 features)
  • Skills: Multiclass classification, one-hot encoding, neural networks
  • Metrics: Accuracy, confusion matrix, visualization

TP3 - Stochastic Training

SVM with SGD on text classification (sci.med vs sci.space newsgroups).

  • Dataset: 20newsgroups (binary classification)
  • Skills: Text vectorization (TF-IDF), SVM with hinge loss, SGD
  • Key Methods: Stochastic gradient descent, grid search

TP4 - Kernels for Text

Kernel SVM for SMS spam detection using spectrum kernels.

  • Datasets: train.csv, val.csv, test.csv (SMS spam data)
  • Skills: Kernel methods, dual SVM formulation, spectrum kernels (unigram/bigram)
  • Key Methods: CountVectorizer, RBF kernel, CVXPY solver, precision/recall/F1

Quick Start

Prerequisites

  • Python 3.10+
  • pip or conda

Installation

# Clone repository
git clone https://github.com/ELFAKIR-taha/ML.git
cd ML

# Create virtual environment
python -m venv venv_vscode
source venv_vscode/Scripts/activate  # Windows
# source venv_vscode/bin/activate    # macOS/Linux

# Install dependencies
pip install -r requirements.txt

Running Labs

VS Code (Recommended)

  1. Open VS Code
  2. Select the venv_vscode interpreter
  3. Open any .ipynb file in TP1-4 folders
  4. Run cells directly in the notebook editor

Jupyter Notebook

jupyter notebook
# Navigate to TP1-4 folders and open .ipynb files

Google Colab (Free GPU)

  1. Upload COLAB_SETUP.ipynb to Google Colab
  2. Mount Google Drive (for data persistence)
  3. All 4 labs + GPU detection built-in

Documentation

Technologies Used

  • PyTorch 2.12.0 - Neural networks & GPU acceleration
  • scikit-learn 1.8.0 - Ridge/Lasso, SVM, vectorizers
  • NumPy 2.4.6 - Numerical computations
  • Pandas 3.0.3 - Data loading & manipulation
  • Matplotlib & Seaborn - Visualization
  • CVXPY 1.9.1 - Convex optimization (dual SVM)
  • SciPy 1.17.1 - Sparse matrices & scientific computing

Features

  • Automatic package installation (setup cells in each notebook)
  • GPU detection & CUDA support
  • Google Colab integration
  • Comprehensive documentation
  • Real datasets (superconductor, penguins, 20newsgroups, SMS spam)
  • Professional visualizations

System Requirements

  • Windows 10+ / macOS / Linux
  • 4GB RAM minimum (8GB recommended)
  • GPU (optional) - NVIDIA CUDA-enabled GPU for acceleration

Troubleshooting

Import errors?

  • Setup cells auto-install packages, but restart kernel if needed

GPU not detected?

  • Check CUDA installation or use Google Colab (free GPU)

Long file paths on Windows?

  • Add repo to shorter path like C:\ML\

License

MIT License - See LICENSE file

Author

Taha EL FAKIR
Computer Vision student at Télécom Saint-Étienne
Machine Learning Course (S8 FISE2)


For questions or issues, open an issue on GitHub.

About

Machine Learning TP exercises (Ridge, SVM, Classification, Kernels)

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages