Advanced Data Science
Hands-on Foundations of Deep Learning
A lower-intensity path from tensors and gradients to modern neural networks, with a project tutorial and two full presentation weeks.
Tensors and Computational Linear Algebra
How can shape and operations preserve—or change—the meaning of data?
Autograd and Gradient Descent
Which way should a parameter change?
Linear Regression
How does a familiar model learn in PyTorch?
Training and Generalization
When does fitting become useful learning?
Logits and Softmax
How do class scores become probabilities?
Classification Workflow
Which errors matter for the decision?
MLPs and Activations
Why do hidden layers need nonlinearity?
PBL Tutorial
How does a question become a defensible data story?
Convolution and CNN Workflow
Why scan local neighborhoods, and how do we test the result?
PBL 1 Presentations
Present evidence, answer questions, and learn from peers.
RNNs and LSTMs
How can a model carry and control earlier information?
Attention and Transformers
How can positions compare directly?
Pretraining and Fine-Tuning
When should representations be reused?
PBL 2 Presentations
Defend the final project and synthesize the course.