PUC-SP Machine Learning Course Resource Hub: A Complete Learning Path from PyTorch to GANs

The PUC-SP Machine Learning Course Resource Hub is the main repository for the 5th semester Machine Learning course at Pontifical Catholic University of São Paulo (PUC-SP) in 2026, maintained by Professor Roney Coelho and hosted on GitHub. This hub offers a structured learning path from PyTorch fundamentals to advanced GANs, covering core content such as deep learning basics, CNNs, RNNs/LSTMs, and Generative Adversarial Networks (GANs). Its features include dual-framework learning (PyTorch and TensorFlow), hands-on projects (image classification, text sentiment analysis, GAN generation), and social practice projects (applying AI to social issues like agriculture, healthcare, and education), combining theoretical depth, practical operability, and the educational concept of social responsibility.

Course Background

Pontifical Catholic University of São Paulo (PUC-SP) is a top-tier higher education institution in Brazil, and its Machine Learning course is characterized by an equal emphasis on theory and practice.

Course Positioning

This course is at an intermediate to advanced level, assuming learners have prior experience in Python programming, linear algebra and calculus, basic statistics, and Pandas/NumPy data processing.

Course Structure

Organized by week, core topics include:

1. Weeks 1-4: Deep Learning Basics (Introduction to PyTorch, Neural Network Fundamentals, Training Techniques, Regularization)
2. Weeks 5-8: Convolutional Neural Networks (CNN Architectures, Image Classification Projects, Transfer Learning)
3. Weeks 9-12: Recurrent Neural Networks (RNN/LSTM, Text Generation and Sentiment Analysis)
4. Weeks 13-16: Generative Adversarial Networks (GAN Principles, DCGAN, Conditional GANs)

Core Methods

• Dual-framework learning: Covers both PyTorch (dynamic computation graph, object-oriented design) and TensorFlow (static graph, Keras API)
• Practical projects: Weekly programming assignments (image classification, text analysis, GAN generation) + social practice projects (AI solving social issues)

Supporting Resources

• Jupyter Notebooks: Interactive tutorials and assignments
• Slides: Theoretical lecture PPTs
• Reading List: Weekly recommended papers
• Video Lectures: Recorded explanatory videos
• Discussion Forum: Student Q&A community

Development Environment

• Google Colab (free GPU), Kaggle Notebooks, local Anaconda environment

Auxiliary Tools

• Weights & Biases (experiment tracking), Hugging Face (pre-trained models), Papers with Code (paper-code comparison)

Self-Learner Path

1. Preparatory Stage (1-2 weeks): Solidify Python, NumPy/Pandas, and math fundamentals
2. Deep Learning Introduction (3-4 weeks): PyTorch/TensorFlow basics, fully connected network implementation
3. Computer Vision (4-5 weeks): CNN theory and practice, image classification projects
4. Natural Language Processing (4-5 weeks): RNN/LSTM, text projects
5. Generative Models (3-4 weeks): GAN principles, DCGAN implementation
6. Comprehensive Project: End-to-end project development

Time Commitment

10-15 hours per week, total duration of about 16 weeks (4 months), with an additional 2-4 weeks for project practice

Community Collaboration

Learning groups, code reviews, and team projects are encouraged, and open-source contributions (such as translation and material improvement) are welcome.

Resource Comparison

Resource	Features	Target Audience
PUC-SP Hub	Structured curriculum, practical projects, social responsibility	Students seeking systematic learning
Fast.ai	Top-down approach, quick to get started	Learners focused on quick application
CS231n	In-depth theory, academic-oriented	Students with math background
deeplearning.ai	Comprehensive and systematic, certificate certification	Career development-oriented
Andrew Ng's ML Course	Classic introduction, clear explanations	Beginners with no prior experience

Limitations and Improvements

• Limitations: Original materials may be in Portuguese, datasets are Western-biased, GPU hardware requirements exist, update frequency needs improvement
• Improvement directions: Add multilingual support, diversify datasets, cloud labs, industry collaboration

The PUC-SP Machine Learning Course Resource Hub represents the open-source trend of high-quality AI education, opening university-level systematic courses to global learners and breaking geographical and economic barriers. It not only provides learning content that combines theory and practice but also emphasizes the social responsibility application of AI technology, making it suitable for students and self-learners who wish to systematically study deep learning. Whether you are a PUC-SP student or a global self-learner, this hub can provide valuable guidance for your AI learning journey.