1. Wrap up
Congratulations! And thank you for completing this course on dimensionality reduction. Let's summarize the journey.
2. Chapter 1 - Dimensionality reduction, feature information
In chapter one, we introduced dimensionality reduction and gained an intuition for feature information through the lenses of missing values, variance, and correlation.
We calculated information gain in decision trees to gain an intuition for measuring feature importance.
Then we discussed the curse of dimensionality, data sparsity, and overfitting.
3. Chapter 2 - Unsupervised feature selection
In chapter two, we differentiated feature selection from feature extraction and performed unsupervised feature selection using missing value ratios, low-variance cutoffs, and correlation filters.
For each of those, we learned how to use the appropriate tidymodels recipe step.
4. Chapter 3 - Supervised feature selection
In chapter three, we reviewed model building with tidymodels and then fit lasso regression and random forest models — two supervised feature selection techniques. We learned how each performs natural feature selection and how to evaluate the reduced model performance.
5. Chapter 4 - Feature extraction
In the last chapter on feature extraction, we learned about principal components and how feature vectors contribute to them. Then we performed and visualized feature extraction using PCA, t-SNE, and UMAP and incorporated PCA and UMAP into tidymodel workflows.
6. Congratulations!
Again, thank you and congratulations! Now, go forth and use these tools to battle the curse of dimensionality!