🧠 Learning Neural Networks, Deep Learning, and LLMs with Python

This repository contains a practical and progressive roadmap to learn from the fundamentals of neural networks to large language models (LLMs), using Python, TensorFlow, and PyTorch, with executable examples on a laptop and open datasets.

---

📌 General Index

🔰 Module 1: Fundamentals of Neural Networks

Objective: Understand and implement basic neural networks from scratch.

• 1.0 Brief review: linear algebra and calculus for neural networks (optional)
• 1.1 Perceptron and learning rule (NumPy)
  • Dataset: Iris (UCI)
• 1.2 Simple Multilayer Perceptron (MLP)
  • Dataset: MNIST
• 1.3 Manual backpropagation
• 1.4 Optimization: SGD, Adam, and other optimizers
• 1.5 First MLP with TensorFlow and PyTorch

---

🔁 Module 2: Deep Neural Networks (DNN)

Objective: Train deep networks and apply improvement techniques.

• 2.1 Regularization: Dropout and L2
  • Dataset: Fashion MNIST
• 2.2 Validation and hyperparameter tuning
• 2.3 Techniques to avoid overfitting (BatchNorm, augmentation)
• 2.4 Introduction to MLOps: model and experiment versioning (for exploration)
• 2.5 Project: Digit Classifier with GUI (Tkinter)

---

📷 Module 3: Convolutional Neural Networks (CNN)

Objective: Apply deep learning to computer vision.

• 3.1 CNN Fundamentals (convolutions, pooling, etc.)
  • Dataset: CIFAR-10
• 3.2 Classic architectures: LeNet, VGG, ResNet
• 3.3 Transfer Learning (MobileNet/ResNet)
  • Dataset: Oxford Flowers, Dogs vs Cats
• 3.4 Modern architectures: EfficientNet, Vision Transformers (for exploration)
• 3.5 Project: Real-time classifier with webcam (OpenCV)

---

🧾 Module 4: NLP and Recurrent Networks

Objective: Process text and sequential data.

• 4.1 Tokenization and embeddings (Word2Vec, GloVe)
  • Dataset: IMDB, Amazon Reviews
• 4.2 LSTM/RNN for text generation
  • Dataset: Recipes, Shakespeare
• 4.3 Text classification with RNN/CNN
• 4.4 Introduction to transformers for NLP (for exploration)
• 4.5 Project: Basic ChatBot (intents + NLP)

---

🧠 Module 5: Transformers and Language Models (LLMs)

Objective: Introduce modern architectures for NLP.

• 5.1 Attention mechanism and Transformer architecture
• 5.2 BERT vs GPT comparison and downstream tasks
  • Dataset: SST2, CoNLL-2003
• 5.3 Fine-tuning with small HuggingFace models
• 5.4 Introduction to GPT-2 and text generation
• 5.5 Project: Educational text generator (fine-tune GPT-2)

---

🔍 Module 6: Interpretability and Evaluation

Objective: Understand model decisions.

• 6.1 Visualization of filters and activations (CNN)
• 6.2 Interpretability with Grad-CAM and LIME
• 6.3 SHAP and LIME for text models (NLP)
• 6.4 Advanced metrics: F1, confusion matrix

---

🚀 Module 7: Production and Optimization

Objective: Take models to production and make them efficient.

• 7.1 Quantization and pruning (TensorFlow Lite / ONNX)
• 7.2 Export models to mobile/web (TF.js, CoreML)
• 7.3 Distributed training, GPU usage (Colab / local)
• 7.4 Basic MLOps practices (for exploration)
• 7.5 Final project: App with embedded model (classifier or generator)

---

📚 Recommended Resources and Datasets

• HuggingFace Datasets
• Kaggle Datasets
• UCI Machine Learning Repository
• OpenML

🔧 Frameworks and Tools

• PyTorch, TensorFlow, Keras, scikit-learn
• transformers, datasets (HuggingFace)
• matplotlib, seaborn, OpenCV, Gradio, Streamlit