TY - JOUR
T1 - Ta2O5-memristor synaptic array with winner-take-all method for neuromorphic pattern matching
AU - Truong, Son Ngoc
AU - Van Pham, Khoa
AU - Yang, Wonsun
AU - Min, Kyeong Sik
AU - Abbas, Yawar
AU - Kang, Chi Jung
AU - Shin, Sangho
AU - Pedrotti, Ken
N1 - Publisher Copyright:
© 2016, The Korean Physical Society.
PY - 2016/8/1
Y1 - 2016/8/1
N2 - Pattern matching or pattern recognition is one of the elemental components that constitute the very complicated recalling and remembering process in human’s brain. To realize this neuromorphic pattern matching, we fabricated and tested a 3 × 3 memristor synaptic array with the winner-take-all method in this research. In the measurement, first, the 3 × 3 Ta2O5 memristor array is programmed to store [LLL], [LHH], and [HLH], where L is a low-resistance state and H is a high-resistance state, at the 1st, 2nd, and 3rd columns, respectively. After the programming, three input patterns, [111], [100], and [010], are applied to the memristor synaptic array. From the measurement results, we confirm that all three input patterns can be recognized well by using a twin memristor crossbar with synaptic arrays. This measurement can be thought of as the first real verification of the twin memristor crossbar with memristive synaptic arrays for neuromorphic pattern recognition.
AB - Pattern matching or pattern recognition is one of the elemental components that constitute the very complicated recalling and remembering process in human’s brain. To realize this neuromorphic pattern matching, we fabricated and tested a 3 × 3 memristor synaptic array with the winner-take-all method in this research. In the measurement, first, the 3 × 3 Ta2O5 memristor array is programmed to store [LLL], [LHH], and [HLH], where L is a low-resistance state and H is a high-resistance state, at the 1st, 2nd, and 3rd columns, respectively. After the programming, three input patterns, [111], [100], and [010], are applied to the memristor synaptic array. From the measurement results, we confirm that all three input patterns can be recognized well by using a twin memristor crossbar with synaptic arrays. This measurement can be thought of as the first real verification of the twin memristor crossbar with memristive synaptic arrays for neuromorphic pattern recognition.
UR - http://www.scopus.com/inward/record.url?scp=84983684733&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84983684733&partnerID=8YFLogxK
U2 - 10.3938/jkps.69.640
DO - 10.3938/jkps.69.640
M3 - Article
AN - SCOPUS:84983684733
SN - 0374-4884
VL - 69
SP - 640
EP - 646
JO - Journal of the Korean Physical Society
JF - Journal of the Korean Physical Society
IS - 4
ER -