Deep Learning & AI Methods

A curated collection of Jupyter notebooks featuring implementations of foundational and advanced algorithms in deep learning, reinforcement learning, and probabilistic modeling.

Overview

This repository serves as a comprehensive reference for core AI/ML techniques, with implementations built from first principles using NumPy and PyTorch. Each module includes detailed explanations, mathematical derivations, and practical examples.

Repository Structure

Directory	Description
deep_learning/	Neural network architectures including autoencoders, RNNs, LSTMs, and dimensionality reduction
probabilistic_models/	Bayesian networks, inference algorithms, and parameter/structure learning
reinforcement_learning/	RL algorithms from bandits to policy gradients and deep RL

Topics Covered

Deep Learning

• Autoencoders & Variational Autoencoders (VAE)
• Recurrent Neural Networks (RNN) & LSTM
• Real-Time Recurrent Learning (RTRL)
• t-SNE Dimensionality Reduction

Probabilistic Models

• Bayesian Network Inference
• Hidden Markov Models (HMM)
• Approximate Inference (Rejection Sampling, Likelihood Weighting)
• Parameter & Structure Learning

Reinforcement Learning

• Multi-Armed Bandits
• Dynamic Programming (Value/Policy Iteration)
• Monte Carlo Methods
• Temporal Difference Learning (SARSA, Q-Learning)
• Deep Q-Networks (DQN)
• Policy Gradient Methods & PPO
• Imitation Learning (DAgger)

Requirements

• Python 3.8+
• PyTorch
• NumPy
• Jupyter Notebook
• Additional dependencies as specified in individual notebooks

Usage

Clone the repository and navigate to the desired topic:

git clone https://github.com/username/deep_learning.git
cd deep_learning
jupyter notebook

License

This project is licensed under the MIT License. See LICENSE for details.