SignRFF: Sign Random Fourier Features

Authors: Xiaoyun Li, Ping Li

NeurIPS 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In addition, experiments are conducted to compare Sign RFF with a wide range of data-dependent and deep learning based hashing methods and show the advantage of Sign RFF with a sufficient number of hash bits. We conduct experiments to demonstrate the effectiveness of our approach and justify that ranking efficiency indeed provides reliable prediction of the empirical search accuracy.
Researcher Affiliation Industry Xiaoyun Li, Ping Li Linked In Ads 700 Bellevue WA NE, Bellevue, WA 98004, USA
Pseudocode No The paper does not contain any structured pseudocode or algorithm blocks.
Open Source Code No The paper states "Did you include the code, data, and instructions needed to reproduce the main experimental results? [Yes] Public datasets." but does not provide a concrete link or explicit instruction on how to access the code.
Open Datasets Yes We use three popular benchmark datasets for image retrieval. The SIFT dataset [23] contains 1M 128dimensional SIFT image features, and 1000 query samples. The MNIST dataset [29] contains 60000 hand-written digits. The CIFAR dataset [25] contains 50000 natural images and we use the gray-scale images in our experiments.
Dataset Splits No The paper mentions using a "test set" for queries but does not explicitly provide train/validation/test splits (e.g., percentages or counts) for all datasets used for training models.
Hardware Specification Yes Our experiments are performed on a single core 2.0GHz CPU.
Software Dependencies No The paper mentions using "VGG-16" for features but does not provide specific version numbers for any software dependencies, libraries, or programming languages used in the experiments.
Experiment Setup Yes For methods (5)-(7) involving Gaussian kernel, we tune γ on a fine grid over 0.1-5 and report the best result.