09_matrix_factorization.md

Section 09: Matrix factorization

Matrix factorization enables us to represent sparse or high-dimensional data sets and high cardinality features with a small number of dense of numeric features suitable for modeling and visualization.

Class Notes

• Basic PCA examples

  • One component with back-projection

  • Iris data with visualization

  • Kaggle House Prices example notebook

• Advanced notes

Supplementary References

• Generalized Low Rank Models (GLRM) with H2O

• LibFM for Factorization Machines

---

• Elements of Statistical Learning
  Sections 14.5 - 14.6, 14.8

• Pattern Recognition in Machine Learning
  Chapter 12

---

Generalized Low Rank Models (GLRM)

• Generalized Low Rank Models (Book)
  by Madeleine Udell, Corinne Horn, Reza Zadeh, and Stephen Boyd

• Generalized Low Rank Models (Paper)
  by Madeleine Udell, Corinne Horn, Reza Zadeh, and Stephen Boyd

• Learning the Parts of Objects by Nonnegative Matrix Factorization
  by Daniel D. Lee and H. Sebastian Seung

• Sparse Principal Component Analysis
  by Hui Zou, Trevor Hastie, and Robert Tibshirani

• Robust Principal Component Analysis?
  by Emmanuel J. Candes, Xiaodong Li, Yi Ma, and John Wright

---

• Factorization Machines
  by Steffen Rendle

• Near Optimal Signal Recovery From Random Projections: Universal Encoding Strategies?
  by Emmanuel Candes and Terence Tao

---

• SAS random projections example

• Quora answer regarding feature extraction