practical-machine-learning

A collection of hands-on machine learning projects focused on implementing and analyzing core algorithms using scientific Python tools.

The repository explores different learning paradigms through practical experiments in clustering, regression, and classification, with an emphasis on data analysis, visualization, optimization, and model evaluation.

---

Projects

Unsupervised Learning

Implementation and exploration of unsupervised learning techniques for discovering hidden structures in data.

Covered topics

• K-Means clustering
• Gaussian Mixture Models (GMM)
• Principal Component Analysis (PCA)
• Latent space visualization
• Cluster evaluation metrics

Datasets

• French cities climate dataset
  (monthly temperatures, geographical information)
• Handwritten digits dataset
  (image clustering and dimensionality reduction tasks)

---

Regression

Experiments focused on predictive modeling for continuous variables using linear and regularized methods.

Covered topics

• Linear Regression
• Ridge Regression
• Lasso Regression
• Gradient descent optimization
• Temporal feature engineering
• Performance evaluation on time-series signals

Dataset

• Brain-Computer Interface (BCI) Competition IV — Dataset 4
  Electrocorticography (ECoG) signals recorded from epileptic patients performing finger movements.
  The objective is to predict thumb flexion from multichannel neural activity recorded at high frequency.

---

Classification

Implementation and comparison of supervised classification algorithms on structured and image datasets.

Covered topics

• Linear Discriminant Analysis (LDA)
• Quadratic Discriminant Analysis (QDA)
• Logistic Regression
• Support Vector Machines (SVM)
• k-Nearest Neighbors (k-NN)
• Neural Networks (MLP)
• Random Forests and Gradient Boosting
• Decision boundary visualization
• Confusion matrix and robustness analysis

Datasets

• Pima Indians Diabetes Dataset
  Medical diagnostic dataset used for diabetes prediction from clinical measurements.
• Handwritten Digits Dataset
  Subset of MNIST containing grayscale digit images for multiclass classification tasks.

---

Libraires

• Python
• NumPy
• SciPy
• scikit-learn
• Matplotlib
• Jupyter Notebook

---

Repository Goals

• Strengthen practical machine learning skills
• Implement and analyze core algorithms
• Build reproducible experimentation workflows
• Improve understanding of optimization and evaluation techniques
• Practice scientific computing and data visualization