Abstract
A sign-linear one-bit map from the d-dimensional sphere Sd to the N-dimensional Hamming cube HN = {-1,+1}n is given by × ↦ {sign(x · zj): 1 ≤ j ≤ N}, where {zj} C Sd. For 0 < δ < 1, we estimate N(d,δ), the smallest integer N so that there is a sign-linear map which has the δ-restricted isometric property, where we impose the normalized geodesic distance on Sd and the Hamming metric on HN. Up to a polylogarithmic factor, N(d,δ) ≈ δ-2+2/(d+1), which has a dimensional correction in the power of δ. This is a question that arises from the one-bit sensing literature, and the method of proof follows from geometric discrepancy theory. We also obtain an analogue of the Stolarsky invariance principle for this situation, which implies that minimizing the L2-average of the embedding error is equivalent to minimizing the discrete energy Σi,j (1/2 - d(zi, zj)2, where d is the normalized geodesic distance.
Original language | English (US) |
---|---|
Pages (from-to) | 744-763 |
Number of pages | 20 |
Journal | Sbornik Mathematics |
Volume | 208 |
Issue number | 6 |
DOIs | |
State | Published - 2017 |
Bibliographical note
Publisher Copyright:© 2017 Russian Academy of Sciences (DoM), London Mathematical Society, Turpion Ltd.
Keywords
- Discrepancy
- One-bit sensing
- Restricted isometry property
- Stolarsky principle