Binary Embedding: Fundamental Limits and Fast Algorithm

Authors: Xinyang Yi, Constantine Caramanis, Eric Price

ICML 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our theoretical findings are supported through experiments on both synthetic and real data sets.
Researcher Affiliation Academia Xinyang Yi YIXY@UTEXAS.EDU Constantine Caramanis CONSTANTINE@UTEXAS.EDU Department of Electrical and Computer Engineering, The University of Texas at Austin, Austin, TX 78712 Eric Price ECPRICE@CS.UTEXAS.EDU Department of Computer Science, The University of Texas at Austin, Austin, TX 78712
Pseudocode Yes Algorithm 1 Uniform Random Projection, Algorithm 2 Fast Binary Embedding, and Algorithm 3 Alternative Fast Binary Embedding are all present in the paper.
Open Source Code No The paper does not provide an explicit statement about making its source code open, nor does it provide a link to a code repository for its methodology.
Open Datasets Yes A popular application of binary embedding is image retrieval, as considered in (Gong & Lazebnik, 2011; Gong et al., 2013; Yu et al., 2014). We thus conduct an experiment on the Flickr-25600 dataset that consists of 10k images from Internet.
Dataset Splits No The paper describes using 500 randomly sampled images as query points and the rest as a base for retrieval, but it does not specify explicit training, validation, or test dataset splits with percentages or sample counts.
Hardware Specification No The paper discusses computational complexity but does not provide any specific details about the hardware (e.g., CPU, GPU models, or memory) used for running the experiments.
Software Dependencies No The paper does not mention any specific software dependencies, such as libraries or solvers, along with their version numbers.
Experiment Setup Yes Algorithm FBE needs parameters n, B, which are intermediate dimension and number of blocks respectively. Based on Theorem 3.8, n is required to be proportional to m (up to some logarithmic factors) and B is required to be proportional to log N. We thus set n 1.3m, B 1.8 log N. We also set n 1.3m for FBE-2. In addition, we fix p = 512.