Heterogeneous density-based clustering with a dual-functional memristive array.

Area of Science:

• * Computational Science and Engineering
• * Materials Science and Engineering
• * Data Science

Background:

• * The increasing volume and complexity of big data necessitate advanced clustering techniques.
• * Existing methods struggle with datasets exhibiting heterogeneous densities and varying distributions.
• * Memristor technology offers potential for hardware acceleration of data processing tasks.

Purpose of the Study:

• * To propose a novel dual-mode memristor crossbar array-based algorithm for data clustering.
• * To address the challenge of clustering datasets with heterogeneous densities.
• * To demonstrate the algorithm's efficiency and feasibility for real-world applications.

Main Methods:

• * Utilized a Ta/HfO2/RuO2 memristor array operating in analog and digital modes.
• * Integrated the local outlier factor (LOF) for heterogeneous density handling.
• * Performed parallel Euclidean and K-distance calculations in analog mode; outlier exclusion and clustering in digital mode.

Main Results:

• * The proposed algorithm achieves linear time complexity.
• * Demonstrated significant improvements over representative density-based algorithms on synthetic datasets.
• * Successfully clustered single-molecule localization microscopy data, validating real-world applicability.

Conclusions:

• * The dual-mode memristor array provides an efficient hardware solution for data clustering.
• * The algorithm effectively handles heterogeneous data densities, outperforming existing methods.
• * The approach shows promise for accelerating analysis in fields like microscopy and big data analytics.