Week 2: Classification, PCA, and Intro to PyTorch

This week will center on classification problems using scikit-learn. Students will also learn about PCA as a tool for visualizing high-dimensional data. They’ll be introduced to workflows using both sklearn.pipeline and Prefect. By the end of the week, they'll take their first steps with PyTorch: installing it, connecting to the GPU, and using it as a replacement for NumPy by manipulating arrays (tensors).

Focus

Train classification models with scikit-learn
Evaluate performance with accuracy, precision, recall, and F1
Use PCA to reduce data to 2D for visualization
Structure ML code using scikit-learn Pipelines and Prefect Flows
Begin working with tensors using PyTorch, learn about the power of GPU.

Hands-On Activities

Train kNN, logistic regression, and decision tree models
Generate and interpret a confusion matrix
Visualize a higher-dimensional dataset using PCA
Build a Pipeline using sklearn.pipeline to chain preprocessing and modeling
Use Prefect to structure a simple ML workflow
Install PyTorch locally.
Use PyTorch to create and manipulate tensors

Learning Outcomes

By the end of this week, students will be able to:

Train and evaluate multiple classification models
Visualize and understand a confusion matrix
Interpret accuracy, precision, recall, and F1 score
Understand the purpose of ML workflows and use basic tools like Pipelines and Prefect Flows
Use PCA to visualize classification boundaries in 2D
Describe what a tensor is and use PyTorch to explore tensor operations
Recognize tell-tale signs of overfitting in classification

Resources

Instructor Notes

The main focus will be on building intuitions with traditional classification tasks and metrics (i.e, not neural networks or tensors), and implementing workflows with pipelines. We will introduce PyTorch at the end just to make sure students can install it, and use it as a vehicle like NumPy for manipulating arrays (tensors).