Parallelisasi Hybrid PSO-GA Menggunakan Island Model dan Cellular Model untuk Optimasi Fungsi pada Multicore CPU dan GPU

Putra, Fehru Mandala (2026) Parallelisasi Hybrid PSO-GA Menggunakan Island Model dan Cellular Model untuk Optimasi Fungsi pada Multicore CPU dan GPU. Other thesis, Institut Teknologi Sepuluh Nopember.

Text skripsi-main_scan.pdf - Accepted Version Restricted to Repository staff only Download (43MB) | Request a copy

Abstract

Particle Swarm Optimization (PSO) dan Genetic Algorithm (GA) telah terbukti efektif dalam menyelesaikan masalah optimasi kompleks. Hibridisasi kedua algoritma dalam bentuk Hybrid PSO-GA menunjukkan performa unggul dibandingkan algoritma tunggal, terutama pada masalah berdimensi tinggi. Namun, implementasi Hybrid PSO-GA yang ada masih bersifat sequential dan belum memanfaatkan arsitektur komputasi paralel modern seperti multicore CPU dan GPU. Penelitian ini mengembangkan dua paradigma parallelisasi Hybrid PSO-GA: island model untuk multicore CPU dan cellular model untuk GPU, dengan tujuan meningkatkan efisiensi komputasi sambil menjaga kualitas solusi tetap kompetitif. Island model membagi populasi menjadi subpopulasi independen dengan migrasi periodik menggunakan ring topology, sementara cellular model mengorganisasikan individu dalam grid 2D dengan interaksi berbasis neighborhood yang cocok untuk paralelisme masif GPU. Evaluasi dilakukan pada dua belas fungsi CEC2022 benchmark suite (F1–F12, mencakup fungsi dasar, hybrid, dan komposisi) pada dimensi 10, 20, 50, dan 100, dengan metrik kualitas solusi, performa komputasi (execution time, speedup, efisiensi paralel), dan skalabilitas. Hasil menunjukkan bahwa keunggulan speedup bersifat spesifik terhadap algoritma dan dimensi: cellular model paling cepat pada dimensi rendah, sedangkan Island-PSO unggul pada dimensi tinggi, dengan titik silang di sekitar dimensi 18–20. Pada anggaran evaluasi setara, kualitas solusi antar arsitektur umumnya sebanding, dan arsitektur cellular terbukti mencegah velocity explosion PSO pada fungsi multimodal berdimensi tinggi.
=====================================================================================================================================
Particle Swarm Optimization (PSO) and Genetic Algorithm (GA) have proven effective in solving complex optimization problems. The hybridization of both algorithms in the form of Hybrid PSO-GA demonstrates superior performance compared to standalone algorithms, especially on high-dimensional problems. However, existing Hybrid PSO-GA implementations remain sequential and have not exploited modern parallel computing architectures such as multicore CPUs and GPUs. This research develops two parallelization paradigms for Hybrid PSO-GA: an island model for multicore CPUs and a cellular model for GPUs, aiming to improve computational efficiency while keeping solution quality competitive. The island model divides the population into independent subpopulations with periodic migration using a ring topology, while the cellular model organizes individuals in a 2D grid with neighborhood-based interactions suitable for massive GPU parallelism. Evaluation is conducted on twelve CEC2022 benchmark suite functions (F1–F12, covering basic, hybrid, and composition functions) at dimensions of 10, 20, 50, and 100, using metrics of solution quality, computational performance (execution time, speedup, parallel efficiency), and scalability. The results show that the speedup advantage is specific to the algorithm and dimensionality: the cellular model is fastest at low dimensions, whereas Island-PSO dominates at high dimensions, with a crossover around dimension 18–20. Under an equal evaluation budget, solution quality across architectures is generally comparable, and the cellular architecture is shown to prevent PSO velocity explosion on high-dimensional multimodal functions.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Hybrid PSO-GA, Komputasi Paralel, Island Model, Cellular Model, CEC2022, Hybrid PSO-GA, Parallel Computing, Island Model, Cellular Model, CEC2022
Subjects:	Q Science > QA Mathematics
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Mathematics > 44201-(S1) Undergraduate Thesis
Depositing User:	Fehru Mandala Putra
Date Deposited:	03 Aug 2026 02:58
Last Modified:	03 Aug 2026 02:58
URI:	http://repository.its.ac.id/id/eprint/142135

Actions (login required)

View Item