Course: Introduction to Deep Learning Focus: Practical implementation of neural networks and deep learning algorithms using Python

This repository contains a comprehensive collection of Jupyter notebooks, assignments, and practice implementations covering fundamental to advanced deep learning concepts. All code is written in Python using industry-standard frameworks including TensorFlow and Keras.

This repository demonstrates mastery of:

✅ Neural Network Fundamentals - Perceptrons, activation functions, backpropagation
✅ Deep Learning Architectures - CNNs, RNNs, and specialized networks
✅ Gradient Descent Optimization - SGD, Adam, RMSprop, learning rate scheduling
✅ TensorFlow & Keras - Model building, training, and deployment
✅ Computer Vision - Image classification, feature extraction, transfer learning
✅ Model Evaluation - Training/validation splits, performance metrics, overfitting prevention

deep-learning-python/
├── assets/                         # Handwritten solutions for assignments and lectures
│   ├── HW1/                        # Assets for assignment1.ipynb
│   ├── HW2/                        # Assets for assignment2.ipynb
│   ├── ...
│   ├── HW6/                        # Assets for assignment6.ipynb
│   ├── Lecture2/                   # Assets for lecture2.ipynb
│   └── Lecture6/                   # Assets for lecture6a/b/c/d.ipynb
├── Assignments/                    # Python code for Deep Learning
│   ├── assignment1-test.ipynb      # Test script for Assignment 1
│   ├── assignment1.ipynb           # Code for Assignment 1
│   ├── assignment2-test.ipynb      # Test script for Assignment 2
│   ├── assignment2.ipynb           # Code for Assignment 2
│   ├── ...
│   └── assignment6.ipynb           # Code for Assignment 6
├── Lectures/                       # Lecture notebooks and examples
│   ├── lecture1.ipynb              # Code from 1st week of lectures
│   ├── lecture1.ipynb              # Code from 2nd week of lectures
│   ├── ...
│   └── lecture7e.ipynb             # Code from 7th week of lectures
├── PracticeExams/                  # Exam prep materials
│   ├── 3dplotTest.ipynb/           # 3D rendering script for GPU testing
│   ├── practiceExam1-11.ipynb      # Midterm practice problems
│   ├── practiceExam1-12.ipynb      # Midterm practice problems
│   ├── practiceExam1-13.ipynb      # Midterm practice problems
│   ├── practiceExam1-14.ipynb      # Midterm practice problems
│   ├── practiceExam1-15.ipynb      # Midterm practice problems
│   └── quiz5.ipynb                 # Practice quiz question
├── assets/                         # Images, diagrams, and resources
└── README.md                       # This document

1. Perceptron Algorithm

• Single-layer perceptrons
• Linear separability
• Decision boundaries
• Weight updates and bias

2. Neural Networks

• Multi-layer perceptrons (MLPs)
• Activation functions (ReLU, sigmoid, tanh, softmax)
• Forward propagation
• Backpropagation algorithm

3. Gradient Descent

• Batch gradient descent
• Stochastic gradient descent (SGD)
• Mini-batch gradient descent
• Momentum and adaptive learning rates

4. Convolutional Neural Networks (CNNs)

• Convolution layers and kernels
• Pooling operations (max, average)
• Feature maps and filters
• Image classification tasks

5. Deep Learning Techniques

• Dropout regularization
• Batch normalization
• Transfer learning
• Data augmentation

6. Model Optimization

• Loss functions (MSE, cross-entropy)
• Optimizers (Adam, RMSprop, Adagrad)
• Learning rate scheduling
• Early stopping

Python Environment:

• Python 3.8 or higher
• Mambaforge package manager (recommended)
• Conda/Mamba environments

Hardware Setup:

• Development Machine: Dell Precision 5540 Laptop
  • Intel Core i9 processor
  • NVIDIA Quadro T2000 (4GB VRAM) - GPU acceleration for model training
  • CUDA-enabled TensorFlow for local GPU training
  • 16GB+ system RAM recommended
  • SSD storage for faster data loading

💡 GPU Advantage: All models in this repository were trained using the NVIDIA Quadro T2000, significantly reducing training time compared to CPU-only execution. TensorFlow automatically detects and utilizes the GPU when properly configured.

Using Mambaforge (Recommended):

# Clone the repository
git clone https://github.com/ADolbyB/deep-learning-python.git
cd deep-learning-python

# Create conda environment with Python 3.10
mamba create -n deep-learning python=3.10
mamba activate deep-learning

# Install TensorFlow with GPU support
mamba install -c conda-forge tensorflow-gpu cudatoolkit cudnn

# Install additional packages
mamba install -c conda-forge keras numpy matplotlib jupyter scikit-learn pandas

# Verify GPU detection
python -c "import tensorflow as tf; print('GPU Available:', tf.config.list_physical_devices('GPU'))"

# Launch Jupyter Notebook or VS Code
jupyter notebook
# Or use VS Code with Jupyter extension

Environment Location:

• Conda environments stored at: ~/mambaforge/envs/deep-learning/
• Package cache: ~/mambaforge/pkgs/

VS Code Setup (GPU-Accelerated Development):

1. Install VS Code extensions:

   • Python
   • Jupyter
   • Pylance

2. Select the conda environment:

   • Press Ctrl+Shift+P
   • Type "Python: Select Interpreter"
   • Choose ~/mambaforge/envs/deep-learning/bin/python

3. Open any .ipynb notebook and run cells with GPU acceleration

🎯 Pro Tip: Use watch -n 1 nvidia-smi in a separate terminal to monitor GPU utilization during training.

1. Navigate to Lectures - Start with Lectures/ for fundamentals
2. Work through Assignments - Assignments/ are structured in order to follow assignments
3. Review Practice Exams - Test and modify to understand concepts
4. Experiment - Modify code and explore different approaches

Perceptron Implementation

• From-scratch perceptron algorithm
• Visualization of decision boundaries
• Binary classification problems

Neural Network Training

• Multi-layer network construction
• Custom training loops
• Performance evaluation and metrics

CNN Image Classification

• Image preprocessing pipelines
• Convolutional layer design
• Transfer learning with pre-trained models

Course: Introduction to Deep Learning
Level: Upper-division Computer Science Elective
Format: Jupyter Notebooks with embedded explanations and visualizations

Learning Approach:

• Theory combined with practical implementation
• Progressive difficulty from fundamentals to advanced topics
• Real-world datasets and problems
• Emphasis on understanding why algorithms work, not just how

• TensorFlow Tutorials
• Keras Getting Started Guide
• Deep Learning Specialization - Coursera

• "Deep Learning" by Ian Goodfellow, Yoshua Bengio, Aaron Courville
• "Hands-On Machine Learning" by Aurélien Géron
• "Neural Networks and Deep Learning" by Michael Nielsen (free online)

• 3Blue1Brown: Neural Networks
• Sentdex: Deep Learning with Python
• Stanford CS231n: CNNs for Visual Recognition

Code Organization:

• ✅ Modular, reusable functions
• ✅ Clear variable naming and documentation
• ✅ Proper train/validation/test splits
• ✅ Reproducible results with random seeds

Model Development:

• ✅ Baseline model establishment
• ✅ Iterative improvement and experimentation
• ✅ Hyperparameter tuning
• ✅ Performance visualization and analysis

Documentation:

• ✅ Markdown cells explaining concepts
• ✅ Inline comments for complex operations
• ✅ Visualizations of results and metrics
• ✅ Lessons learned and insights

While this is primarily a coursework repository, improvements are welcome:

• 📝 Documentation enhancements
• 🐛 Bug fixes in implementations
• 💡 Additional examples or explanations
• 🎨 Visualization improvements

Please open an issue or submit a pull request!

This project is licensed under the GNU GPL v3 License - see the LICENSE.md file for details.

Academic Integrity Notice: This repository represents completed coursework. If you're currently enrolled in a similar course, please use this as reference material only and adhere to your institution's academic honesty policies.

GitHub: Joel Brigida
LinkedIn: Joel Brigida

Questions about implementations or concepts? Feel free to open an issue!