01. Random Projection

[Note: this lesson comes after a lesson explaining Principal Component Analysis]

Paper: Random projection in dimensionality reduction: Applications to image and text data

This paper examines using Random Projection to reduce the dimensionality of image and text data. It shows how Random Projection proves to be a computationally simple method of dimensionality reduction, while still preserving the similarities of data vectors to a high degree. The paper shows this on real-world datasets including noisy and noiseless images of natural scenes, and text documents from a newsgroup corpus.

Paper: Random Projections for k-means Clustering

This paper uses Random Projection as an efficient dimensionality reduction step before conducting k-means clustering on a dataset of 400 face images of dimensions 64 × 64.