Accelerating Random Kaczmarz Algorithm Based on Clustering Information

Authors: Yujun Li, Kaichun Mo, Haishan Ye

AAAI 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental After applying clustering method and utilizing clustering information, we improve RKA-JL and RKA-Block algorithms to speedup their convergences. The empirical experiments show the improvement clearly. In section 4, we conduct some numerical experiments to show the improved performance of our algorithms.
Researcher Affiliation Academia Yujun Li and Kaichun Mo and Haishan Ye Department of Computer Science and Engineering Shanghai Jiao Tong University {liyujun145,yhs12354123}@gmail.com, daerduomkch@sjtu.edu.cn
Pseudocode Yes Algorithm 1 RKA-JL, Algorithm 2 RKA-Block, Algorithm 3 RKA-Cluster-JL, Algorithm 4 RKA-Cluster-Block
Open Source Code No The paper does not provide any statement about releasing source code or links to a code repository for the described methodology.
Open Datasets No The paper states: "We generate data that comprises of several clusters." and "We generate data that lies in four distinctive clusters." This indicates synthetic data generation rather than the use of a publicly available dataset with specific access information or a formal citation.
Dataset Splits No The paper mentions generating data and adding noise, but it does not specify explicit training/validation/test splits, percentages, or sample counts needed for data partitioning.
Hardware Specification Yes The experiments are run in a PC with WIN7 system, i5-3470 core 3.2GHz CPU and 8G RAM.
Software Dependencies No The paper mentions using a "K-means variant algorithm (King 2012)" and refers to "Matlab notation", but it does not specify version numbers for any software, libraries, or programming languages used for implementation, which is required for reproducibility.
Experiment Setup Yes First, we compare the proposed RKA-Cluster-JL with the original RKA-JL algorithm. We generate data that comprises of several clusters. Here, n = 10000 and p = 1000. Besides, since the real data is usually corrupted by white noise, we add Gaussian noise with mean 0 and standard deviation 0.1 or 0.2. Next, we compare the results produced by RKA-Block and RKA-Cluster-Block algorithms. We generate data that lies in four distinctive clusters. Here, n = 10000, p = 1000 and the block size is four. Then, as usual, white Gaussian noise with mean 0 and standard deviation 0.1 or 0.2 is added to the data to simulate the real world.