Minsuk Koo

Prof. Minsuk Koo

orcid

Research Focus & Metrics

Academic Career Length

8Years (2019-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
28 June 2021 in IEEE Solid-State Circuits Letters
Verified
Journal Article
Closed access
AAAmogh Agrawal
Amogh Agrawal
MAMustafa Ali
Mustafa Ali
MKMinsuk Koo
Minsuk Koo
NRNitin Rathi
Nitin Rathi
Akhilesh JaiswalAkhilesh Jaiswal
Akhilesh Jaiswal
Kaushik RoyKaushik Roy
Kaushik Roy
The inherent dynamics of the neuron membrane potential in spiking neural networks (SNNs) allows the processing of sequential learning tasks, avoiding the complexity of recurrent neural networks. The highly sparse spike-based computations in such spatiotemporal data can be leveraged for energy efficiency. However, the membrane potential incurs additional memory access bottlenecks in current SNN hardware. To that effect, we propose a 10T-SRAM compute-in-memory (CIM) macro, specifically designed for state-of-the-art SNN inference. It consists of a fused weight ( W MEM ) and membrane potential ( V MEM ) memory and inherently exploits sparsity in input spikes leading to ~97.4% reduction in energy-delay product (EDP) at 85% sparsity (typical of SNNs considered in this work) compared to the case of no sparsity. We propose staggered data mapping and reconfigurable peripherals for handling different bit precision requirements of W MEM and V MEM , while supporting multiple neuron functionalities. The proposed macro was fabricated in 65-nm CMOS technology, achieving energy efficiency of 0.99 TOPS/W at 0.85-V supply and 200-MHz frequency for signed 11-bit operations. We evaluate the SNN for sentiment classification from the IMDB dataset of movie reviews and achieve within ~1% accuracy difference and ~ 5× higher energy efficiency compared to a corresponding long short-term memory network.

16 March 2020 in IEEE Transactions on Circuits and Systems I: Regular Papers
Verified
Journal Article
Closed access
MKMinsuk Koo
Minsuk Koo
GSGopalakrishnan Srinivasan
Gopalakrishnan Srinivasan
YSYong Shim
Yong Shim
Kaushik RoyKaushik Roy
Kaushik Roy
In this work, we propose stochastic Binary Spiking Neural Network (sBSNN) composed of stochastic spiking neurons and binary synapses (stochastic only during training) that computes probabilistically with one-bit precision for power-efficient and memory-compressed neuromorphic computing. We present an energy-efficient implementation of the proposed sBSNN using `stochastic bit' as the core computational primitive to realize the stochastic neurons and synapses, which are fabricated in 90nm CMOS process, to achieve efficient on-chip training and inference for image recognition tasks. The measured data shows that the `stochastic bit' can be programmed to mimic spiking neurons, and stochastic Spike Timing Dependent Plasticity (or sSTDP) rule for training the binary synaptic weights without expensive random number generators. Our results indicate that the proposed sBSNN realization offers possibility of up to 32x neuronal and synaptic memory compression compared to full precision (32-bit) SNN and energy efficiency of 89.49 TOPS/Watt for two-layer fully-connected SNN.

1 September 2020 in Physical Review Applied
Verified
Journal Article
Closed access
PDF file
MKMinsuk Koo
Minsuk Koo
MPM.R. Pufall
M.R. Pufall
YSYong Shim
Yong Shim
AKA.B. Kos
A.B. Kos
GCGyorgy Csaba
Gyorgy Csaba
Kaushik RoyKaushik Roy
Kaushik Roy
Recent research on nano-oscillators has shown the possibility of using a coupled-oscillator network as a core-computing primitive for non-Boolean computation. The spin-torque oscillator (STO) is an attractive candidate because it is CMOS compatible, highly integrable, scalable, and frequency and phase tunable. Based on these promising features, we propose an alternative coupled-oscillator-based architecture for hybrid spintronic and CMOS hardware that computes a multidimensional norm. The hybrid system, composed of an array of four injection-locked STOs and a CMOS detector, is experimentally demonstrated. The measured performance is then used as the input to simulations that demonstrate the hybrid system as both a distance metric and a convolution computational primitive for image-processing applications. Energy and scaling analysis shows that the STO-based coupled-oscillatory system has a higher efficiency than the CMOS-based system with an order of magnitude faster computation speed in distance computation for high-dimensional input vectors.

23 November 2023 in Advanced Intelligent Systems
Verified
Journal Article
Open access
JKJaesung Kim
Jaesung Kim
JKJung Nam Kim
Jung Nam Kim
YKYoon Kim
Yoon Kim
SHSungmin Hwang
Sungmin Hwang
MKMinsuk Koo
Minsuk Koo
Spiking neural networks (SNNs) have been researched as an alternative to reduce the gap with the human brain in terms of energy efficiency, due to their inherent spare event‐driven characteristics from a hardware implementation perspective. However, they still face significant challenges in learning, compared to artificial neural networks (ANNs). Recently, several algorithms have been developed to narrow the performance gap between SNNs and ANNs, including features in spiking neurons that can reduce information loss in the membrane potential. Inspired by these advancements, the current study designs and measures a neuron circuit using 180 nm complementary metal‐oxide‐semiconductor (CMOS) technology to address this information loss. The proposed circuit successfully implements these features, and their performance is validated through simulation based on the measured data.

4 October 2021 in IEEE Electron Device Letters
Verified
Journal Article
Closed access
EYEunseon Yu
Eunseon Yu
AAAmogh Agrawal
Amogh Agrawal
DZDongqi Zheng
Dongqi Zheng
MSMengwei Si
Mengwei Si
MKMinsuk Koo
Minsuk Koo
Kaushik RoyKaushik Roy
Kaushik Roy
Coupled-oscillatory networks are an emerging paradigm for efficiently solving optimization problems. In this work, we demonstrate the application of ferroelectric field-effect transistor (FeFET) for energy-efficient coupled-oscillatory networks. A CMOS-compatible FeFET was fabricated having > 1V of hysteresis window and 57 mV/dec of minimum subthreshold swing. With our proposed FeFET oscillator circuits and optimized biasing schemes, a $2\times $ wider synchronization range and up to $276\times $ lower energy per cycle were achieved compared to previous FeFET-based oscillators. Moreover, we employ FeFET coupled-oscillatory network for an edge detection task. Our simulations considering FeFET non-idealities and process variations with a 5-bit quantized image show that the edge detection output closely follows the ideal output.

23 September 2024 in Biomimetics
Verified
Journal Article
Open access
PDF file
GKGyeongpoy Kim
Gyeongpoy Kim
SPSeoyoung Park
Seoyoung Park
MKMinsuk Koo
Minsuk Koo
Sungjun KimSungjun Kim
Sungjun Kim
In this study, we investigate the impact of O2 plasma treatment on the performance of Al/TaOX/Al-based resistive random-access memory (RRAM) devices, focusing on applications in neuromorphic systems. Comparative analysis using scanning electron microscopy and X-ray photoelectron spectroscopy confirmed the differences in chemical composition between O2-plasma-treated and untreated RRAM cells. Direct-current measurements showed that O2-plasma-treated RRAM cells exhibited significant improvements over untreated RRAM cells, including higher on/off ratios, improved uniformity and distribution, longer retention times, and enhanced durability. The conduction mechanism is investigated by current–voltage (I–V) curve fitting. In addition, paired-pulse facilitation (PPF) is observed using partial short-term memory. Furthermore, 3- and 4-bit weight tuning with auto-pulse-tuning algorithms was achieved to improve the controllability of the synapse weight for the neuromorphic system, maintaining retention times exceeding 103 s in the multiple states. Neuromorphic simulation with an MNIST dataset is conducted to evaluate the synaptic device.