Hybrid Random Projection: Integrating Dense and Sparse Techniques for Enhanced Representation in High-Dimensional Data

29 Pages Posted: 24 Jan 2024

See all articles by Yussif Yahaya

Yussif Yahaya

affiliation not provided to SSRN

Jimoh Olawale Ajadi

affiliation not provided to SSRN

Ridwan A. Sanusi

affiliation not provided to SSRN

Zaid Sawlan

affiliation not provided to SSRN

nurudeen adegoke

Massey University

Abstract

This paper introduces a new hybrid random projection method (HBP) that effectively combines the strengths of normal random projection (NRP) and plus-minus one random projection (PMRP). By incorporating a blending parameter, HBP optimises the contributions of NRP and PMRP, improving the process of reducing the data dimensions while ensuring that the data structure is preserved. Traditional methods such as NRP and PMRP are known for their benefits, but they also have limitations. NRP is generally accurate but can struggle with certain data types or noise. PMRP is simple but sometimes fails to capture complex relationships within the data. The effectiveness of HBP is studied through detailed evaluations, including simulations and real-world data analyses from key machine learning datasets, such as period change, toxicity, MNIST, and human activity recognition (HAR). We examine various factors such as sample size, dimensions of the original and reduced data, and sparsity. The primary metric used to measure the performance is distance distortion, which indicates how well the structure of the dataset is maintained after dimension reduction. In every test, HBP consistently shows the least distance distortion, performing better than both NRP and PMRP. This is true for a range of scenarios and datasets. HBP represents a significant advancement in dimension reduction techniques. Its ability to minimise the loss of information during the reduction process demonstrates its potential as a powerful tool for managing complex high-dimensional data in practical applications. This makes HBP an important development in the fields of machine learning, intelligent systems and artificial intelligence, offering improved efficiency and precision in data analyses.

Keywords: High-dimensional data analysis, dimensionality reduction techniques, hybrid random projection, normal random projection, plus-minus one random projection.

Suggested Citation

Yahaya, Yussif and Ajadi, Jimoh Olawale and Sanusi, Ridwan A. and Sawlan, Zaid and adegoke, nurudeen, Hybrid Random Projection: Integrating Dense and Sparse Techniques for Enhanced Representation in High-Dimensional Data. Available at SSRN: https://ssrn.com/abstract=4705621 or http://dx.doi.org/10.2139/ssrn.4705621

Yussif Yahaya

affiliation not provided to SSRN ( email )

No Address Available

Jimoh Olawale Ajadi

affiliation not provided to SSRN ( email )

No Address Available

Ridwan A. Sanusi

affiliation not provided to SSRN ( email )

No Address Available

Zaid Sawlan

affiliation not provided to SSRN ( email )

No Address Available

Nurudeen Adegoke (Contact Author)

Massey University ( email )

Private Bag 11 222
Palmerston North, 4442
New Zealand

Do you have negative results from your research you’d like to share?

Paper statistics

Downloads
21
Abstract Views
66
PlumX Metrics