Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

High-Performance Hybrid Algorithm for Minimum Sum-of-Squares Clustering of Infinitely Tall Data

Ravil Mussabayev, Rustam Mussabayev

TL;DR

Evaluation of HPClust, an innovative set of hybrid parallel approaches for the minimum sum-of-squares clustering of infinitely tall data, demonstrates its superiority over traditional and cutting-edge methods by offering better performance in the key metrics.

Abstract

This paper introduces a novel formulation of the clustering problem, namely the Minimum Sum-of-Squares Clustering of Infinitely Tall Data (MSSC-ITD), and presents HPClust, an innovative set of hybrid parallel approaches for its effective solution. By utilizing modern high-performance computing techniques, HPClust enhances key clustering metrics: effectiveness, computational efficiency, and scalability. In contrast to vanilla data parallelism, which only accelerates processing time through the MapReduce framework, our approach unlocks superior performance by leveraging the multi-strategy competitive-cooperative parallelism and intricate properties of the objective function landscape. Unlike other available algorithms that struggle to scale, our algorithm is inherently parallel in nature, improving solution quality through increased scalability and parallelism, and outperforming even advanced algorithms designed for small and medium-sized datasets. Our evaluation of HPClust, featuring four parallel strategies, demonstrates its superiority over traditional and cutting-edge methods by offering better performance in the key metrics. These results also show that parallel processing not only enhances the clustering efficiency, but the accuracy as well. Additionally, we explore the balance between computational efficiency and clustering quality, providing insights into optimal parallel strategies based on dataset specifics and resource availability. This research advances our understanding of parallelism in clustering algorithms, demonstrating that a judicious hybridization of advanced parallel approaches yields optimal results for MSSC-ITD. Experiments on synthetic data further confirm HPClust's exceptional scalability and robustness to noise.

High-Performance Hybrid Algorithm for Minimum Sum-of-Squares Clustering of Infinitely Tall Data

TL;DR

Evaluation of HPClust, an innovative set of hybrid parallel approaches for the minimum sum-of-squares clustering of infinitely tall data, demonstrates its superiority over traditional and cutting-edge methods by offering better performance in the key metrics.

Abstract

This paper introduces a novel formulation of the clustering problem, namely the Minimum Sum-of-Squares Clustering of Infinitely Tall Data (MSSC-ITD), and presents HPClust, an innovative set of hybrid parallel approaches for its effective solution. By utilizing modern high-performance computing techniques, HPClust enhances key clustering metrics: effectiveness, computational efficiency, and scalability. In contrast to vanilla data parallelism, which only accelerates processing time through the MapReduce framework, our approach unlocks superior performance by leveraging the multi-strategy competitive-cooperative parallelism and intricate properties of the objective function landscape. Unlike other available algorithms that struggle to scale, our algorithm is inherently parallel in nature, improving solution quality through increased scalability and parallelism, and outperforming even advanced algorithms designed for small and medium-sized datasets. Our evaluation of HPClust, featuring four parallel strategies, demonstrates its superiority over traditional and cutting-edge methods by offering better performance in the key metrics. These results also show that parallel processing not only enhances the clustering efficiency, but the accuracy as well. Additionally, we explore the balance between computational efficiency and clustering quality, providing insights into optimal parallel strategies based on dataset specifics and resource availability. This research advances our understanding of parallelism in clustering algorithms, demonstrating that a judicious hybridization of advanced parallel approaches yields optimal results for MSSC-ITD. Experiments on synthetic data further confirm HPClust's exceptional scalability and robustness to noise.
Paper Structure (47 sections, 1 equation, 7 figures, 30 tables, 5 algorithms)

This paper contains 47 sections, 1 equation, 7 figures, 30 tables, 5 algorithms.

Table of Contents

  1. Introduction
  2. Related Works
  3. Proposed Algorithm
  4. Parallel Strategies for the HPClust Algorithm
  5. High-Performance Techniques in HPClust
  6. Analytical optimization
  7. Nuances of parallelism in HPClust
  8. Implementation-level optimization
  9. Future optimization directions
  10. Experimental Setup
  11. Hardware and software
  12. Competitive algorithms
  13. Datasets
  14. Experimental design and evaluation metrics
  15. Hyperparameter selection
  16. ...and 32 more sections

Figures (7)

  • Figure S1: Flowchart of the HPClust algorithm with the competitive parallelism
  • Figure S2: Flowchart of the HPClust algorithm using a cooperative parallel strategy
  • Figure S3: Comparative results of the algorithms with respect to the number of employed CPUs averaged across all datasets
  • Figure S4: Comparative results of the algorithms with respect to the number of points $m$ in a synthetic dataset
  • Figure A5: Number of distance evaluations, 1
  • ...and 2 more figures