서브메뉴
검색
Emulating Advanced Neural Functions With Spin Textures for Neuromorphic Computing
Emulating Advanced Neural Functions With Spin Textures for Neuromorphic Computing
상세정보
- 자료유형
- 학위논문 서양
- 최종처리일시
- 20250211153109
- ISBN
- 9798384442394
- DDC
- 621.3
- 저자명
- Cui, Can.
- 서명/저자
- Emulating Advanced Neural Functions With Spin Textures for Neuromorphic Computing
- 발행사항
- [Sl] : The University of Texas at Austin, 2024
- 발행사항
- Ann Arbor : ProQuest Dissertations & Theses, 2024
- 형태사항
- 122 p
- 주기사항
- Source: Dissertations Abstracts International, Volume: 86-04, Section: B.
- 주기사항
- Advisor: Incorvia, Jean Anne.
- 학위논문주기
- Thesis (Ph.D.)--The University of Texas at Austin, 2024.
- 초록/해제
- 요약Neuromorphic computing calls for devices that exhibit intrinsic neuronal functions for compact, energy efficient hardware implementations. Spintronic devices allow for electrical or magnetic control of the spin polarization of the computing media and are gaining importance in neuromorphic computing due to benefits in endurance, energy efficiency and CMOS processing compatibility. Most importantly, their reconfigurable spin textures provide additional flexibility in mimicking the complex, dynamic behaviors of the biological neurons and synapses.This dissertation explores the mapping of neuronal functions onto spin textures within ferromagnetic thin films, particularly focusing on domain walls (DW) and skyrmions. Through a combination of device modeling and experimental studies, I aim to demonstrate that the manipulability of these spin textures holds significant promise for neuromorphic computing applications. I will introduce the fundamental concepts of magnetism and magnetic materials that lay the groundwork for understanding DW and skyrmion dynamics. The integrated DW-magnetic tunnel junction (MTJ) device is first discussed as a basis for spintronic neuromorphic applications. Device-level micromagnetic modeling reveals the mechanism behind the intrinsic lateral inhibition (LI) phenomena between DW-MTJ integrate-fire neurons, and I will show the methodology to maximize LI, which is essential for regulating neuronal activities and facilitating winner-take-all (WTA) functions. A novel adaptive neuron is designed by leveraging the rich dynamics of skyrmions to emulate advanced cognitive functions such as adaptive responses and cross-frequency coupling (CFC). Device functionality is validated using micromagnetic modeling, and their superiority in context-aware learning is demonstrated through neural network modeling.The experimental study focuses on the design and fabrication of a DW-MTJ integrate-fire neuron device which improves the DW injection and reset characteristics of previous implementations. Fabrication techniques and optimization are presented, along with experimental demonstration of device functionality. I highlight the future integration of neuron and synapse devices into a monolithic, all-spintronic neuromorphic network, paving the way for stream learning applications. Overall, this dissertation contributes to advancing the application of spintronic devices in neuromorphic computing, offering insights into the potential of spin textures for next-generation cognitive computing systems.
- 일반주제명
- Computer engineering
- 일반주제명
- Electromagnetics
- 일반주제명
- Condensed matter physics
- 일반주제명
- Information technology
- 기타저자
- The University of Texas at Austin Electrical and Computer Engineering
- 기본자료저록
- Dissertations Abstracts International. 86-04B.
- 전자적 위치 및 접속
- 로그인 후 원문을 볼 수 있습니다.
MARC
008250123s2024 us c eng d■001000017164973
■00520250211153109
■006m o d
■007cr#unu||||||||
■020 ▼a9798384442394
■035 ▼a(MiAaPQ)AAI31690549
■035 ▼a(MiAaPQ)123vireo24677Cui
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a621.3
■1001 ▼aCui, Can.
■24510▼aEmulating Advanced Neural Functions With Spin Textures for Neuromorphic Computing
■260 ▼a[Sl]▼bThe University of Texas at Austin▼c2024
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2024
■300 ▼a122 p
■500 ▼aSource: Dissertations Abstracts International, Volume: 86-04, Section: B.
■500 ▼aAdvisor: Incorvia, Jean Anne.
■5021 ▼aThesis (Ph.D.)--The University of Texas at Austin, 2024.
■520 ▼aNeuromorphic computing calls for devices that exhibit intrinsic neuronal functions for compact, energy efficient hardware implementations. Spintronic devices allow for electrical or magnetic control of the spin polarization of the computing media and are gaining importance in neuromorphic computing due to benefits in endurance, energy efficiency and CMOS processing compatibility. Most importantly, their reconfigurable spin textures provide additional flexibility in mimicking the complex, dynamic behaviors of the biological neurons and synapses.This dissertation explores the mapping of neuronal functions onto spin textures within ferromagnetic thin films, particularly focusing on domain walls (DW) and skyrmions. Through a combination of device modeling and experimental studies, I aim to demonstrate that the manipulability of these spin textures holds significant promise for neuromorphic computing applications. I will introduce the fundamental concepts of magnetism and magnetic materials that lay the groundwork for understanding DW and skyrmion dynamics. The integrated DW-magnetic tunnel junction (MTJ) device is first discussed as a basis for spintronic neuromorphic applications. Device-level micromagnetic modeling reveals the mechanism behind the intrinsic lateral inhibition (LI) phenomena between DW-MTJ integrate-fire neurons, and I will show the methodology to maximize LI, which is essential for regulating neuronal activities and facilitating winner-take-all (WTA) functions. A novel adaptive neuron is designed by leveraging the rich dynamics of skyrmions to emulate advanced cognitive functions such as adaptive responses and cross-frequency coupling (CFC). Device functionality is validated using micromagnetic modeling, and their superiority in context-aware learning is demonstrated through neural network modeling.The experimental study focuses on the design and fabrication of a DW-MTJ integrate-fire neuron device which improves the DW injection and reset characteristics of previous implementations. Fabrication techniques and optimization are presented, along with experimental demonstration of device functionality. I highlight the future integration of neuron and synapse devices into a monolithic, all-spintronic neuromorphic network, paving the way for stream learning applications. Overall, this dissertation contributes to advancing the application of spintronic devices in neuromorphic computing, offering insights into the potential of spin textures for next-generation cognitive computing systems.
■590 ▼aSchool code: 0227.
■650 4▼aComputer engineering
■650 4▼aElectromagnetics
■650 4▼aCondensed matter physics
■650 4▼aInformation technology
■653 ▼aSkyrmion dynamics
■653 ▼aLateral inhibition
■653 ▼aWinner-take-all functions
■653 ▼aSpintronic devices
■653 ▼aSpin polarization
■690 ▼a0489
■690 ▼a0464
■690 ▼a0611
■690 ▼a0607
■71020▼aThe University of Texas at Austin▼bElectrical and Computer Engineering.
■7730 ▼tDissertations Abstracts International▼g86-04B.
■790 ▼a0227
■791 ▼aPh.D.
■792 ▼a2024
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17164973▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.
