Deep Learning for Social Sciences

Topics

This course offers a deep dive into advanced neural network architectures and their applications, including:

• Introduction to the use of deep learning techniques in social sciences
• Basic Machine Learning concepts: regression, classification, and models
• Neural Networks: Perceptrons, Multi-Layer Perceptrons, XOR problem, activation functions, backpropagation, and gradient descent
• Convolutional Neural Networks: Image processing implementations in PyTorch
• Graph Neural Networks and their Python implementations
• Recurrent Neural Networks and Long Short Term Memory networks for time series analysis
• Generative deep learning techniques
• Advanced topics: Diffusion Models, Reinforcement Learning, and Large Language Models based on transformer architectures like the GPT series in NLP with practical implementations in PyTorch

Format

The course is structured as:

1. 12 lectures on the theory of Deep Learning
2. 8 coding labs
3. 1 guest researchers seminars
4. 1 students presentation day

Grading

The course grade is composed of three assignments delivered during the semester (40%) and a final project (60%).

Assignments

You will have 3 weeks to work individually on each assignment. For the second assignment you will have 4 weeks instead. The assignments will be delivered on the following dates:

1. April 30 to May 20 - Multilayer Perceptron
2. May 21 to June 17 - Convolutional Neural Network
3. June 18 to July 9 - Large Language Models

Exercises

A practical part of coding exercises connected to the lectures will be offered. The exercises will not be graded.

Theory Seminars

Labs and Coding Sessions

Seminars and Presentations

Books

• Understanding Deep Learning by Simon J.D. Prince
• The Little Book of Deep Learning by François Fleuret
• Deep Learning Course, a thorough introduction to deep-learning, with examples in the PyTorch framework
• Alice’s Adventures in a differentiable wonderland by Simone Scardapane

Previous assignments examples

• Multilayer Perceptron
• Convolutional Neural Network
• Convolutional Neural Network

Previous projects examples

• CNN for Detecting Deforestation and Ice Melting
• GNN to Detect Fake News
• Using RNNs to Forecast Migration Patterns
• Synthetic Product Review Generation for Market Analysis
• Gentrification Detection from Street View Images
• Income Inequality Prediction Through Country Similarity Networks
• Environmental Risk Discourse Evolution in SEC 10-K Filings

Lessons

Assignments ▾ Final Projects ▾