TY - JOUR
T1 - An Overflow/Underflow-Free Fixed-Point Bit-Width Optimization Method for OS-ELM Digital Circuit
AU - Tsukada, Mineto
AU - Matsutani, Hiroki
N1 - Funding Information:
Manuscript received March 15, 2021. Manuscript revised July 13, 2021. Manuscript publicized September 17, 2021. †The authors are with Graduate School of Science and Technology, Keio University, Yokohama-shi, 223-8522 Japan. ∗This work was partially supported by JST CREST Grant Number JPMJCR20F2, Japan. a) E-mail: tsukada@arc.ics.keio.ac.jp b) E-mail: matutani@arc.ics.keio.ac.jp DOI: 10.1587/transfun.2021VLP0017
Publisher Copyright:
© 2022 Institute of Electronics, Information and Communication, Engineers, IEICE. All rights reserved.
PY - 2022
Y1 - 2022
N2 - Currently there has been increasing demand for real-time training on resource-limited IoT devices such as smart sensors, which realizes standalone online adaptation for streaming data without data transfers to remote servers. OS-ELM (Online Sequential Extreme Learning Machine) has been one of promising neural-network-based online algorithms for on-chip learning because it can perform online training at low computational cost and is easy to implement as a digital circuit. Existing OS-ELM digital circuits employ fixed-point data format and the bit-widths are often manually tuned, however, this may cause overflow or underflow which can lead to unexpected behavior of the circuit. For on-chip learning systems, an overflow/underflow-free design has a great impact since online training is continuously performed and the intervals of intermediate variables will dynamically change as time goes by. In this paper, we propose an overflow/underflow-free bit-width optimization method for fixed-point digital circuits of OS-ELM. Experimental results show that our method realizes overflow/underflow-free OS-ELM digital circuits with 1.0x - 1.5x more area cost compared to the baseline simulation method where overflow or underflow can happen.
AB - Currently there has been increasing demand for real-time training on resource-limited IoT devices such as smart sensors, which realizes standalone online adaptation for streaming data without data transfers to remote servers. OS-ELM (Online Sequential Extreme Learning Machine) has been one of promising neural-network-based online algorithms for on-chip learning because it can perform online training at low computational cost and is easy to implement as a digital circuit. Existing OS-ELM digital circuits employ fixed-point data format and the bit-widths are often manually tuned, however, this may cause overflow or underflow which can lead to unexpected behavior of the circuit. For on-chip learning systems, an overflow/underflow-free design has a great impact since online training is continuously performed and the intervals of intermediate variables will dynamically change as time goes by. In this paper, we propose an overflow/underflow-free bit-width optimization method for fixed-point digital circuits of OS-ELM. Experimental results show that our method realizes overflow/underflow-free OS-ELM digital circuits with 1.0x - 1.5x more area cost compared to the baseline simulation method where overflow or underflow can happen.
KW - OS-ELM
KW - bit-width optimization
KW - fixed-point design
UR - http://www.scopus.com/inward/record.url?scp=85126628211&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85126628211&partnerID=8YFLogxK
U2 - 10.1587/transfun.2021VLP0017
DO - 10.1587/transfun.2021VLP0017
M3 - Article
AN - SCOPUS:85126628211
SN - 0916-8508
VL - 105
SP - 437
EP - 447
JO - IEICE Transactions on Fundamentals of Electronics, Communications and Computer Sciences
JF - IEICE Transactions on Fundamentals of Electronics, Communications and Computer Sciences
IS - 3
ER -