Phase 3 · Machine learning · Unsupervised
Dimensionality reduction
Visualize and compress high-dimensional data.
Open this lesson Create a free account
Why it matters
Embeddings have hundreds or thousands of dimensions, and you can't look at any of them. Dimensionality reduction squeezes data into 2 dimensions to see it, or into fewer dimensions to store and search it more cheaply. You met the math of PCA in Phase 2. This lesson is about using it, and its nonlinear cousins, without being fooled by the pictures they make.
What you'll be able to do
- Use PCA in scikit-learn to visualize data and to compress features before a model
- Choose how many components to keep from the explained variance and from model scores
- Use t-SNE for pictures of high-dimensional data, and read those pictures correctly
- Connect dimensionality reduction to embeddings: plotting them and shortening them
Before this lesson
Next in this phase: The scikit-learn workflow