IEEE Solid-States Circuits Magazine - Fall 2021 - 76

with CSC architectures and can deliver
up to 3.6 TOPJ which is on par with
related accelerators fabricated in the
same technology. Moreover, CSC architectures,
similar to butterfly networks,
can be categorized as one of the highbandwidth
communication networks
for parallel system interconnections.
Such interconnection is used in the
design of a high-bandwidth router embedded
into the CSC-AP.
Acknowledgments
We would like to thank Tejas Kadale
and Kevin Gubbi from San Francisco
State University for their help and
valuable input on this project.
References
[1] M. Hosseini, M. Horton, H. Paneliya, U.
Kallakuri, H. Homayoun, and T. Mohsenin,
" On the complexity reduction of
dense layers from O (n2) to O (nlogn) with
cyclic sparsely connected layers, " in Proc.
56th ACM/IEEE Des. Autom. Conf. (DAC),
2019, pp. 1-6.
[2] S. Han, J. Pool, J. Tran, and W. J. Dally, " Learning
both weights and connections for efficient
neural network, " in Proc. Adv. Neural
Inf. Proc. Syst., 2015, pp. 1135-1143.
[3] Y. Guo, A. Yao, and Y. Chen, " Dynamic network
surgery for efficient DNNs, " in Proc.
Adv. Neural Inf. Process. Syst., 2016, pp.
1387-1395.
[4] H. Mao et al., " Exploring the regularity of
sparse structure in convolutional neural
networks, " 2017, arXiv:1705.08922.
[5] M. Hosseini, N. K. Manjunath, U. Kallakuri,
V. Chandrareddy, and T. Mohsenin, " Cyclic
sparsely connected architectures for compact
deep convolutional neural networks, "
in Proc. IEEE Trans. Very Large Scale Integration
(VLSI) Syst., 2021, pp. 1-8.
[6] C. Deng, S. Liao, Y. Xie, K. K. Parhi, X.
Qian, and B. Yuan, " PermDNN: Efficient
compressed DNN architecture with permuted
diagonal matrices, " in Proc. 51st
Annu. IEEE/ACM Int. Symp. Microarch.
(MICRO), 2018, pp. 189-202. doi: 10.1109/
MICRO.2018.00024.
[7] C. Ding et al., " CirCNN: Accelerating and
compressing deep neural networks using
block-circulant weight matrices, " in Proc.
50th Annu. IEEE/ACM Int. Symp. Microarch.,
2017, pp. 395-408.
[8] Y.-H. Chen, T. Krishna, J. S. Emer, and V.
Sze, " Eyeriss: An energy-efficient reconfigurable
accelerator for deep convolutional
neural networks, " IEEE J. Solid-State
Circuits, vol. 52, no. 1, pp. 127-138, Jan.
2016. doi: 10.1109/JSSC.2016.2616357.
[9] J. Lee, C. Kim, S. Kang, D. Shin, S. Kim,
and H.-J. Yoo, " UNPU: A 50.6 TOPS/W unified
deep neural network accelerator with
1b-to-16b fully-variable weight bit-precision, "
in Proc. IEEE Int. Solid-State Circuits
Conf. (ISSCC), 2018, pp. 218-220.
[10] Z. Yuan et al., " Sticker: A 0.41-62.1
TOPS/W 8bit neural network processor
with multi-sparsity compatible convolution
arrays and online tuning acceleration
for fully connected layers, " in Proc. IEEE
Symp. VLSI Circuits, 2018, pp. 33-34. doi:
10.1109/VLSIC.2018.8502404.
76
FALL 2021
[11] A. Shiri et al., " Energy-efficient hardware
for language guided reinforcement learning, "
in Proc. Great Lakes Symp. VLSI,
2020, pp. 131-136.
[12] H. Ren et al., " End-to-end scalable and low
power multi-modal CNN for respiratoryrelated
symptoms detection, " in Proc.
IEEE 33rd Int. System-on-Chip Conf. (SOCC)
(SOCC 2020), Sept. 2020, pp. 102-107. doi:
10.1109/SOCC49529.2020.9524755.
[13] B. Prakash, N. Waytowich, A. Ganeshan,
T. Oates, and T. Mohsenin, " Guiding safe
reinforcement learning policies using
structured language constraints, " in Proc.
SafeAI Workshop 34th AAAI Conf. Artif.
Intell., 2020. [Online]. Available: http://
eehpc.csee.umbc.edu/publ icat ions/
pdf/2020/AAAI_RL_Workshop.pdf
[14] A. N. Mazumder et al., " Automatic detection
of respiratory symptoms using a low
power multi-input CNN processor, " IEEE
Des. Test., early access, May 11, 2021. doi:
10.1109/MDAT.2021.3079318.
About the Authors
Morteza Hosseini (hs10@umbc.edu)
received his M.Sc. degree in electrical
engineering from Sharif University of
Technology, Tehran, Iran, in 2012. He
is working toward his Ph.D. degree in
computer engineering at the University
of Maryland, Baltimore County,
Baltimore, Maryland, 21250, USA. His
research interests include optimizing
deep neural networks for energy-efficient
deployment to hardware.
Nitheesh Manjunath (n67@umbc
.edu) received his M.Sc. degree in computer
engineering from the University
of Maryland, Baltimore County, Baltimore,
Maryland, 21250, USA, in 2021.
He is a field-programmable gate array
(FPGA) design engineer at Broadcast
Sports International, LLC. His research
interests include energy-efficient machine
learning as well as edge artificial
intelligence accelerator design and
implementation on FPGAs and application-specific
ICs.
Uttej Kallakuri (ukalla1@umbc
.edu) is working toward his Ph.D. degree
in computer engineering from the University
of Maryland, Baltimore County,
Baltimore, Maryland, 21250, USA, where
he earned his M.Sc. degree in 2019. His
research interests include the design
and implementation of energy-efficient
high-performance machine learning algorithms
on field-programmable gate
arrays and application-specific ICs.
Hamid Mahmoodi (mahmoodi@
sfsu.edu) received his Ph.D. degree in
electrical and computer engineering
IEEE SOLID-STATE CIRCUITS MAGAZINE
from Purdue University, West Lafayette,
Indiana, in 2005. He is a professor of
electrical and computer engineering in
the School of Engineering at San Francisco
State University, San Francisco,
California, 94132, USA. His research interests
include low-power, reliable, and
high-performance circuit design in nanoelectronic
technologies. He has published
more than 100 technical papers
in journals and conferences and holds
five U.S. patents.
Houman Homayoun (hhomayoun
@ucdavis.edu) is an associate professor
in the Department of Electrical and
Computer Engineering at University
of California, Davis, Davis, California,
95616, USA. He is also the director of
the National Science Foundation Center
for Hardware and Embedded Systems
Security and Trust. His research
interests include hardware security
and trust, applied machine learning
and artificial intelligence, and data-intensive
and heterogeneous computing.
He has published more than 200 technical
papers at prestigious conferences
and in journals and directed more
than US$8 million in research funding
from the National Science Foundation,
DARPA, Air Force Research Laboratory,
the National Institute of Standards and
Technology, U.S. Congress, and various
industrial sponsors.
Tinoosh Mohsenin (tinoosh@umbc
.edu) is an associate professor with the
Department of Computer Science and
Electrical Engineering, University of
Maryland Baltimore County, Baltimore,
Maryland, 21250, USA, where she is
also the director of the Energy-Efficient
High-Performance Computing Lab. She
received her Ph.D. from the University
of California, Davis, in 2010 and her
M.Sc. degree from Rice University in
2004, both in electrical and computer
engineering. She has authored or coauthored
more than 130 peer-reviewed
journal and conference publications.
Her research interests include designing
energy-efficient embedded processors
for machine learning and signal
processing, knowledge extraction techniques
for autonomous systems, wearable
smart health monitoring, and
embedded big data computing.
http://eehpc.csee.umbc.edu/publications/pdf/2020/AAAI_RL_Workshop.pdf http://eehpc.csee.umbc.edu/publications/pdf/2020/AAAI_RL_Workshop.pdf http://eehpc.csee.umbc.edu/publications/pdf/2020/AAAI_RL_Workshop.pdf
IEEE Solid-States Circuits Magazine - Fall 2021

Table of Contents for the Digital Edition of IEEE Solid-States Circuits Magazine - Fall 2021
