Matrix Factorization Approaches

Definition

The idea is that we transform a feature matrix \(\mathbf{X}\) into a dictionary and a basis/new representation \(\mathbf{H}\). ie. \(\mathbf{X = HD}\).

Many unsupervised learning algorithms (dimensionality reduction, sparse coding) and semi-supervised learning algorithms (supervised dictionary learning) can be formulated as matrix factorization.

The key is dimensionality reduction and data re-representation.

Techinques in dimensionality reduction include: Principal Component Analysis (PCA) Sparse Coding

Example

References

Finding the dictionary to get latent factors Latent Factors and Factor Analysis