Rare Eclipses in Quantised Random Embeddings of Disjoint Convex Sets: a Matter of Consistency?
(2017) Signal Processing with Adaptive Sparse Structured Representations (SPARS′17) Workshop — Location: Lisbon (5.June.2017)
Files
SPARS2017_Paper_33.pdf
Open Access - Adobe PDF
- 382.43 KB
Details
- Authors
- Abstract
- We study the problem of verifying when two disjoint closed convex sets remain separable after the application of a quantised random embedding, as a means to ensure exact classification from the signatures produced by this non-linear dimensionality reduction. An analysis of the interplay between the embedding, its quantiser resolution and the sets’ separation is presented in the form of a convex problem; this is completed by its numerical exploration in a special case, for which the phase transition corresponding to exact classification is easily computed.
- Affiliations
Citations
Cambareri, V., Xu, C., & Jacques, L. (2017). Rare Eclipses in Quantised Random Embeddings of Disjoint Convex Sets: a Matter of Consistency? Proceedings of SPARS′17. Published. Signal Processing with Adaptive Sparse Structured Representations (SPARS′17) Workshop, Lisbon. https://hdl.handle.net/2078.5/227362