Optimal Convergence Rates for Distributed Nystroem Approximation

Authors: Jian Li, Yong Liu, Weiping Wang

JMLR 2023 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Finally, we conduct experiments on several real-world datasets to validate the effectiveness of the proposed algorithm, and the empirical results coincide with our theoretical findings. In this section, we study the generalization performance on real-world datasets. We implement all methods based on Pytorch 1.13 1 and run experiments on a Linux Server with an Nvidia RTX 2080Ti GPU.
Researcher Affiliation Academia Jian Li EMAIL Institute of Information Engineering, Chinese Academy of Sciences, Beijing, China Yong Liu EMAIL Gaoling School of Artificial Intelligence, Renmin University of China, Beijing, China Beijing Key Laboratory of Big Data Management and Analysis Methods, Beijing, China Weiping Wang EMAIL Institute of Information Engineering, Chinese Academy of Sciences, Beijing, China
Pseudocode Yes Algorithm 1 Distributed Nystr om approximation (DNystr om)
Open Source Code Yes 1. Publicly available at https://github.com/superlj666/DNystroem
Open Datasets Yes We evaluate the compared algorithms on real-world classification datasets, which are publicly available from UCI datasets 2. Available at http://archive.ics.uci.edu/ml/datasets.php
Dataset Splits Yes Table 1: The statistics and tuned hyperparameters in datasets usps 10 7291 2007 256 10 10 6 pendigits 10 7494 3498 16 100 10 6 letter 26 15000 5000 16 1 10 7 MNIST 10 60000 10000 784 10 10 6
Hardware Specification Yes We implement all methods based on Pytorch 1.13 1 and run experiments on a Linux Server with an Nvidia RTX 2080Ti GPU.
Software Dependencies Yes We implement all methods based on Pytorch 1.13 1 and run experiments on a Linux Server with an Nvidia RTX 2080Ti GPU.
Experiment Setup Yes Using the toolbox NNI 3, we tune the optimal hyperparameters over the grids σ {10i, i = 4, 3, , 4} and λ {10i, i = 10, , 1}. The statistics information and optimal hyperparameters for datasets are recorded in Table 1.