Genetic Programming Empowered Feature Construction towards Energy Efficient BVI Wearables

Created by W.Langdon from gp-bibliography.bib Revision:1.8051

@InProceedings{xu:2024:GECCO3,

• author = "Peijie Xu and Andy Song and Ke Wang",
• title = "Genetic Programming Empowered Feature Construction towards Energy Efficient {BVI} Wearables",
• booktitle = "Proceedings of the 2024 Genetic and Evolutionary Computation Conference",
• year = "2024",
• editor = "Ruhul Sarker and Patrick Siarry and Julia Handl and Xiaodong Li and Markus Wagner and Mario Garza-Fabre and Kate Smith-Miles and Richard Allmendinger and Ying Bi and Grant Dick and Amir H Gandomi and Marcella Scoczynski Ribeiro Martins and Hirad Assimi and Nadarajen Veerapen and Yuan Sun and Mario Andres Munyoz and Ahmed Kheiri and Nguyen Su and Dhananjay Thiruvady and Andy Song and Frank Neumann and Carla Silva",
• pages = "1408--1416",
• address = "Melbourne, Australia",
• series = "GECCO '24",
• month = "14-18 " # jul,
• organisation = "SIGEVO",
• publisher = "Association for Computing Machinery",
• publisher_address = "New York, NY, USA",
• keywords = "genetic algorithms, genetic programming, obstacle avoidance datasets, obstacle detection, feature construction, low energy consumption, efficient models, hardware deployment, power usage, Real World Applications",
• isbn13 = "979-8-4007-0494-9",
• DOI = "doi:10.1145/3638529.3654002",
• size = "9 pages",
• abstract = "Blind and visually impaired (BVI) individuals face serious mobility-related risks daily due to the lack of progression in hazard detection and assistive technologies. Most existing techniques are demanding in computational resources and energy consumption, yet still struggle in real-time performance. The challenge is particularly evident when deploying these techniques on portable devices, on which lightweight coupled with sustainable low battery usage is a must. Hence in this study, we aim to leverage Genetic Programming (GP), which is well known for its feature construction capability, to develop more energy-efficient models with better features. The objective is to find more condensed features by GP, to reduce energy consumed and inference time, but without significant accuracy loss in obstacle detection from a head-mount wearable device for BVIs. Features have been trained on a series of in-door settings that represent a BVI person navigating through corridors and furniture. Models with these constructed features then are validated on actual portable board deployment, as well as on Field Programmable Gate Array simulation (FPGA). Comparative analyses are conducted using a range of performance metrics across models training by different learning methods. The metrics include accuracy, model execution time, prediction time, energy consumption, and hardware resource usage. Our study demonstrates that GP-constructed models can generally reduce energy consumption and inference time with negligible accuracy loss. Furthermore, it can build models with higher accuracy than the benchmark, allowing users to adjust between energy usage and accuracy according to their needs.",
• notes = "GECCO-2024 RWA A Recombination of the 33rd International Conference on Genetic Algorithms (ICGA) and the 29th Annual Genetic Programming Conference (GP)",

}

Genetic Programming entries for Peijie Xu Andy Song Ke Wang

Citations