Binary Embedding with Additive Homogeneous Kernels

Authors: Saehoon Kim, Seungjin Choi

AAAI 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Numerical experiments demonstrate that BE-AHK actually yields similarity-preserving binary codes in terms of additive homogeneous kernels and is superior to existing methods in case that training data and queries are generated from different distributions.
Researcher Affiliation Academia Saehoon Kim, Seungjin Choi Department of Computer Science and Engineering Pohang University of Science and Technology 77 Cheongam-ro, Nam-gu, Pohang 37673, Korea {kshkawa,seungjin}@postech.ac.kr
Pseudocode No The paper describes the proposed algorithm's steps in paragraph text but does not include a clearly labeled pseudocode or algorithm block.
Open Source Code No The paper does not provide any explicit statements about releasing open-source code for the methodology described.
Open Datasets Yes MNIST is composed of 784-dimensional 50,000 training data and 10,000 testing data with 10 classes. GIST1M (J egou, Douze, and Schmid 2011) is composed of 920-dimensional 1 million GIST descriptors with additional 1,000 queries.
Dataset Splits Yes MNIST is composed of 784-dimensional 50,000 training data and 10,000 testing data with 10 classes.
Hardware Specification No The paper does not explicitly describe the specific hardware (e.g., GPU/CPU models, memory) used to run its experiments.
Software Dependencies No The paper does not provide specific version numbers for software components or libraries used in the experiments.
Experiment Setup Yes BE-AHK (the proposed method) has one hyperparameter to set the number of samples per dimension to construct feature maps. We used three (ten) samples per dimension for χ2 (intersection) kernel. KLSH (Jiang, Que, and Kulis 2015) has two hyperparameters: the number of landmark points and the rank of KPCA. We fixed the number of landmark points to be 1,000. We tested different ranks for KPCA, because it it sensitive to the performance of KLSH.