Binary Embedding: Fundamental Limits and Fast Algorithm

Xinyang Yi, Constantine Caramanis, Eric Price
Proceedings of the 32nd International Conference on Machine Learning, PMLR 37:2162-2170, 2015.

Abstract

Binary embedding is a nonlinear dimension reduction methodology where high dimensional data are embedded into the Hamming cube while preserving the structure of the original space. Specifically, for an arbitrary N distinct points in \mathbbS^p-1, our goal is to encode each point using m-dimensional binary strings such that we can reconstruct their geodesic distance up to δuniform distortion. Existing binary embedding algorithms either lack theoretical guarantees or suffer from running time O(mp). We make three contributions: (1) we establish a lower bound that shows any binary embedding oblivious to the set of points requires m =Ω(\frac1δ^2\logN) bits and a similar lower bound for non-oblivious embeddings into Hamming distance; (2) we propose a novel fast binary embedding algorithm with provably optimal bit complexity m = O(\frac1 δ^2\logN) and near linear running time O(p \log p) whenever \log N ≪δ\sqrtp, with a slightly worse running time for larger \log N; (3) we also provide an analytic result about embedding a general set of points K ⊆\mathbbS^p-1 with even infinite size. Our theoretical findings are supported through experiments on both synthetic and real data sets.

Cite this Paper


BibTeX
@InProceedings{pmlr-v37-yi15, title = {Binary Embedding: Fundamental Limits and Fast Algorithm}, author = {Yi, Xinyang and Caramanis, Constantine and Price, Eric}, booktitle = {Proceedings of the 32nd International Conference on Machine Learning}, pages = {2162--2170}, year = {2015}, editor = {Bach, Francis and Blei, David}, volume = {37}, series = {Proceedings of Machine Learning Research}, address = {Lille, France}, month = {07--09 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v37/yi15.pdf}, url = { http://proceedings.mlr.press/v37/yi15.html }, abstract = {Binary embedding is a nonlinear dimension reduction methodology where high dimensional data are embedded into the Hamming cube while preserving the structure of the original space. Specifically, for an arbitrary N distinct points in \mathbbS^p-1, our goal is to encode each point using m-dimensional binary strings such that we can reconstruct their geodesic distance up to δuniform distortion. Existing binary embedding algorithms either lack theoretical guarantees or suffer from running time O(mp). We make three contributions: (1) we establish a lower bound that shows any binary embedding oblivious to the set of points requires m =Ω(\frac1δ^2\logN) bits and a similar lower bound for non-oblivious embeddings into Hamming distance; (2) we propose a novel fast binary embedding algorithm with provably optimal bit complexity m = O(\frac1 δ^2\logN) and near linear running time O(p \log p) whenever \log N ≪δ\sqrtp, with a slightly worse running time for larger \log N; (3) we also provide an analytic result about embedding a general set of points K ⊆\mathbbS^p-1 with even infinite size. Our theoretical findings are supported through experiments on both synthetic and real data sets.} }
Endnote
%0 Conference Paper %T Binary Embedding: Fundamental Limits and Fast Algorithm %A Xinyang Yi %A Constantine Caramanis %A Eric Price %B Proceedings of the 32nd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2015 %E Francis Bach %E David Blei %F pmlr-v37-yi15 %I PMLR %P 2162--2170 %U http://proceedings.mlr.press/v37/yi15.html %V 37 %X Binary embedding is a nonlinear dimension reduction methodology where high dimensional data are embedded into the Hamming cube while preserving the structure of the original space. Specifically, for an arbitrary N distinct points in \mathbbS^p-1, our goal is to encode each point using m-dimensional binary strings such that we can reconstruct their geodesic distance up to δuniform distortion. Existing binary embedding algorithms either lack theoretical guarantees or suffer from running time O(mp). We make three contributions: (1) we establish a lower bound that shows any binary embedding oblivious to the set of points requires m =Ω(\frac1δ^2\logN) bits and a similar lower bound for non-oblivious embeddings into Hamming distance; (2) we propose a novel fast binary embedding algorithm with provably optimal bit complexity m = O(\frac1 δ^2\logN) and near linear running time O(p \log p) whenever \log N ≪δ\sqrtp, with a slightly worse running time for larger \log N; (3) we also provide an analytic result about embedding a general set of points K ⊆\mathbbS^p-1 with even infinite size. Our theoretical findings are supported through experiments on both synthetic and real data sets.
RIS
TY - CPAPER TI - Binary Embedding: Fundamental Limits and Fast Algorithm AU - Xinyang Yi AU - Constantine Caramanis AU - Eric Price BT - Proceedings of the 32nd International Conference on Machine Learning DA - 2015/06/01 ED - Francis Bach ED - David Blei ID - pmlr-v37-yi15 PB - PMLR DP - Proceedings of Machine Learning Research VL - 37 SP - 2162 EP - 2170 L1 - http://proceedings.mlr.press/v37/yi15.pdf UR - http://proceedings.mlr.press/v37/yi15.html AB - Binary embedding is a nonlinear dimension reduction methodology where high dimensional data are embedded into the Hamming cube while preserving the structure of the original space. Specifically, for an arbitrary N distinct points in \mathbbS^p-1, our goal is to encode each point using m-dimensional binary strings such that we can reconstruct their geodesic distance up to δuniform distortion. Existing binary embedding algorithms either lack theoretical guarantees or suffer from running time O(mp). We make three contributions: (1) we establish a lower bound that shows any binary embedding oblivious to the set of points requires m =Ω(\frac1δ^2\logN) bits and a similar lower bound for non-oblivious embeddings into Hamming distance; (2) we propose a novel fast binary embedding algorithm with provably optimal bit complexity m = O(\frac1 δ^2\logN) and near linear running time O(p \log p) whenever \log N ≪δ\sqrtp, with a slightly worse running time for larger \log N; (3) we also provide an analytic result about embedding a general set of points K ⊆\mathbbS^p-1 with even infinite size. Our theoretical findings are supported through experiments on both synthetic and real data sets. ER -
APA
Yi, X., Caramanis, C. & Price, E.. (2015). Binary Embedding: Fundamental Limits and Fast Algorithm. Proceedings of the 32nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 37:2162-2170 Available from http://proceedings.mlr.press/v37/yi15.html .

Related Material