🤖 Machine Learning & Deep Learning from Scratch

A complete and organized repository with Machine Learning and Deep Learning algorithm implementations from scratch, practical projects, and educational notebooks.

📚 Repository Structure

This repository is organized into three main areas:

🔷 1. Machine Learning (Classical Algorithms)

Implementations of fundamental machine learning algorithms.

📊 Supervised Learning

• Regression

  • Linear Regression (Gradient Descent & Normal Equation)
  • Locally Weighted Linear Regression
  • Softmax Regression

• Classification

  • Logistic Regression (Gradient Descent & Newton's Method)
  • Naive Bayes

🔍 Unsupervised Learning

• Clustering
  • K-Means (C++ implementation)

🎯 Nearest Neighbors

• K-Nearest Neighbors (KNN) - C++ implementation

⚙️ C++ Implementations

Low-level C++ implementations for optimized performance.

---

🔷 2. Deep Learning (Neural Networks and Deep Learning)

🧠 Fundamentals

• Basic neural network implementation
• Layers: Input, Hidden, Output
• Neurons and network architecture

👁️ Computer Vision

Classic CNN Architectures:

• LeNet
• VGG16
• GoogLeNet (Inception)
• ResNet
• EfficientNet

Practical Projects:

• Melanoma Classification
• Skin Cancer Classification
• FP16 Optimization

OpenCV:

• Basic tutorials: filters, colors, contours, edge detection
• Projects: Color detection, Face anonymization
• Various experiments

Utilities:

• Data loaders
• Mean and standard deviation calculation
• Binary Focal Loss
• Helper functions

🎨 Generative Models

• Autoencoders
• Fully Connected GAN
• DCGAN (Deep Convolutional GAN)

💬 NLP (Natural Language Processing)

Fundamentals:

• N-grams
• Data gathering and preprocessing

Sequence Models:

• RNN (Recurrent Neural Networks)
• RNN from Scratch
• LSTM (Long Short-Term Memory)
• Sequence-to-Sequence
• Sequence-to-Sequence with Attention

Transformers:

• Transformers from Scratch
• GPT-1 from Scratch
• Transformers for Translation
• Project: GPT trained on Wikipedia articles

Applications:

• Spam Classifier (Naive Bayes)
• Name Generator
• Datasets and metrics

🎓 Specialized Techniques

• Transfer Learning
• Multi-label Classification
• Transpose Convolution

---

🔷 3. Recommendation Systems

• Popularity-based recommendation system

---

📂 Datasets

All datasets are centralized in the data/ folder:

• images/: MNIST and other images
• tabular/: Iris.csv and other tabular datasets

---

🚀 How to Use

Prerequisites

Python:

pip install numpy pandas matplotlib scikit-learn
pip install torch torchvision
pip install tensorflow
pip install opencv-python
pip install nltk

C++ (for native implementations):

• C++ Compiler (g++, MSVC, etc.)
• CMake (optional)

Running Notebooks

1. Install Jupyter:

pip install jupyter notebook

2. Navigate to desired folder:

cd 2-Deep-Learning/computer_vision/architectures
jupyter notebook

Compiling C++ Code

cd 1-Machine-Learning/cpp_implementations
g++ -std=c++17 -o program main.cpp knn.cpp kmeans.cpp data_handler.cpp
./program

---

📖 Topics by Area

Classical Machine Learning

• Linear and Logistic Regression
• Distance-based algorithms (KNN)
• Probabilistic models (Naive Bayes)
• Clustering (K-Means)

Deep Learning - Computer Vision

• Fundamental CNN architectures
• Transfer Learning
• GANs and generative models
• Medical projects (skin cancer detection)
• OpenCV and image processing

Deep Learning - NLP

• Classical language models (N-grams)
• RNNs and LSTMs
• Attention Mechanisms
• Transformers and GPT
• Practical applications

---

🎯 Repository Goals

1. Educational: Didactic implementations from scratch
2. Practical: Real projects and applications
3. Organized: Clear structure by topics and subtopics
4. Complete: From fundamentals to advanced techniques

---

🛠️ Technologies Used

• Languages: Python, C++
• DL Frameworks: PyTorch, TensorFlow
• Computer Vision: OpenCV
• NLP: NLTK, Transformers
• Data: NumPy, Pandas
• Visualization: Matplotlib, Seaborn

---

📝 Detailed Directory Structure

Machine-Learning-Scratch/
├── data/                                      # Centralized datasets
│   ├── images/                               # MNIST, etc.
│   └── tabular/                              # CSV files
│
├── 1-Machine-Learning/                        # Classical ML
│   ├── supervised/
│   │   ├── regression/                       # Regression algorithms
│   │   ├── classification/                   # Classification algorithms
│   │   └── examples/                         # Practical examples
│   ├── unsupervised/
│   │   ├── clustering/                       # K-means
│   │   └── dimensionality_reduction/         # PCA, etc.
│   ├── nearest_neighbors/                    # KNN
│   └── cpp_implementations/                  # C++ implementations
│
├── 2-Deep-Learning/                           # Deep Learning
│   ├── fundamentals/                         # Basic neural networks
│   │   └── layers/                           # Layer implementations
│   ├── computer_vision/
│   │   ├── architectures/                    # Classic CNNs
│   │   ├── projects/                         # Practical projects
│   │   ├── opencv/                           # OpenCV
│   │   └── utils/                            # Utilities
│   ├── generative_models/                    # GANs, Autoencoders
│   ├── nlp/
│   │   ├── fundamentals/                     # N-grams, preprocessing
│   │   ├── sequence_models/                  # RNN, LSTM
│   │   ├── transformers/                     # Transformers, GPT
│   │   └── applications/                     # Practical projects
│   └── specialized/                          # Specialized techniques
│
├── 3-Recommendation-Systems/                  # Recommendation systems
│
└── README.md                                  # This file

---

🤝 Contributions

This is a personal study repository, but suggestions and improvements are welcome!

---

📄 License

This project is for educational purposes.

---

📧 Contact

• GitHub: @MatheusLevy

---

Last updated: October 2025