Deep Learning for Dynamic Network Analysis (DLDNA)

Course Overview

This graduate-level course introduces deep learning techniques with a focus on Graph Neural Networks (GNNs) for analyzing dynamic networks. Students learn to implement neural network architectures using PyTorch and apply these techniques to engineering problems in transportation and urban systems.

Instructor: Bahman Madadi
Format: 8 sessions including lectures, hands-on practice, and project presentations

Learning Objectives

Upon completion, students will be able to:

  • Differentiate between major deep learning architectures (CNNs, RNNs, Transformers, GNNs) and select appropriate methods for engineering problems
  • Design and implement complete deep learning workflows using PyTorch, including data preparation, model training, validation, and inference
  • Apply Graph Neural Networks to solve engineering problems in network analysis and dynamic systems
  • Evaluate model performance using appropriate metrics and diagnose issues
  • Critically assess capabilities, limitations, and appropriate use cases of deep learning models

Topics Covered

  • Neural Network Fundamentals: Architecture design, training, and optimization
  • Modern Deep Learning: CNNs, RNNs, Transformers, and attention mechanisms
  • Graph Neural Networks: Message passing, graph convolutions, and network learning
  • Advanced Learning Paradigms: Transfer learning, meta-learning, and large language models
  • Practical Implementation: Complete PyTorch workflow from data to deployment
  • Engineering Applications: Transportation networks, urban systems, and infrastructure analysis

Course Structure

The course combines:

  • Theory Sessions: Core concepts and architectures
  • Hands-on Practice: Weekly PyTorch coding exercises (TP)
  • Quizzes: Individual assessments (50% of grade)
  • Course Project: Group project applying deep learning to real-world problems (50% of grade)

Tools & Technologies

  • Primary Framework: PyTorch and PyTorch Geometric
  • Programming: Python with focus on practical implementation

All course materials, including lecture slides and Jupyter notebooks, are provided via Moodle.