Compressive classification and the rare eclipse problem
This work addresses the rare eclipse problem in compressive classification, offering theoretical guarantees for linear separability with few measurements, which is incremental for domains like hyperspectral imaging.
The paper tackles the problem of when convex sets remain separable after random projection, providing bounds for ellipsoids based on distance and polynomial coefficients, and demonstrates that data can be classified with very few measurements while staying linearly separable, as shown in hyperspectral imaging applications.
This paper addresses the fundamental question of when convex sets remain disjoint after random projection. We provide an analysis using ideas from high-dimensional convex geometry. For ellipsoids, we provide a bound in terms of the distance between these ellipsoids and simple functions of their polynomial coefficients. As an application, this theorem provides bounds for compressive classification of convex sets. Rather than assuming that the data to be classified is sparse, our results show that the data can be acquired via very few measurements yet will remain linearly separable. We demonstrate the feasibility of this approach in the context of hyperspectral imaging.