Exploiting Approximate Feature Extraction via Genetic Programming for Hardware Acceleration in a Heterogeneous Microprocessor

Created by W.Langdon from gp-bibliography.bib Revision:1.7615

@Article{Jia:2018:ieeeJSSC,

• author = "Hongyang Jia and Naveen Verma",
• journal = "IEEE Journal of Solid-State Circuits",
• title = "Exploiting Approximate Feature Extraction via Genetic Programming for Hardware Acceleration in a Heterogeneous Microprocessor",
• year = "2018",
• volume = "53",
• number = "4",
• pages = "1016--1027",
• month = apr,
• keywords = "genetic algorithms, genetic programming, Approximate computation, feature extraction, machine learning, programmable accelerator, sensor inference",
• ISSN = "0018-9200",
• URL = "http://www.princeton.edu/~nverma/VermaLabSite/Publications/2018/JiaVerma_JSSC2018.pdf",
• DOI = "doi:10.1109/JSSC.2017.2787762",
• size = "12 pages",
• abstract = "This paper presents a heterogeneous microprocessor for low-energy sensor-inference applications. Hardware acceleration has shown to enable substantial energy-efficiency and throughput gains, but raises significant challenges where programmable computations are required, as in the case of feature extraction. To overcome this, a programmable feature-extraction accelerator (FEA) is presented that exploits genetic programming for automatic program synthesis. This leads to approximate, but highly structured, computations, enabling: 1) a high degree of specialization; 2) systematic mapping of programs to the accelerator; and 3) energy scalability via user-controllable approximation knobs. A microprocessor integrating a CPU with feature-extraction and classification accelerators is prototyped in 130-nm CMOS. Two medical-sensor applications (electroencephalogram-based seizure detection and electrocardiogram-based arrhythmia detection) demonstrate 325times and 156times energy reduction, respectively, for programmable feature extraction implemented on the accelerator versus a CPU-only architecture, and 7.6times and 6.5times energy reduction, respectively, versus a CPU-with-coprocessor architecture. Furthermore, 20times and 9times energy scalability, respectively, is demonstrated via the approximation knobs. The energy-efficiency of the programmable FEA is 220 GOPS/W, near that of fixed-function accelerators in the same technology, exceeding typical programmable accelerators.",
• notes = "Also known as \cite{8262650}",

}

Genetic Programming entries for Hongyang Jia Naveen Verma

Citations