본문

서브메뉴

Machine Learning for Knowledge Discovery in Engineering and Science: From Nanophotonic Design to Medical Diagnosis
Machine Learning for Knowledge Discovery in Engineering and Science: From Nanophotonic Des...
Machine Learning for Knowledge Discovery in Engineering and Science: From Nanophotonic Design to Medical Diagnosis

Detailed Information

자료유형  
 학위논문 서양
최종처리일시  
20260209102907
ISBN  
9798263396695
DDC  
790
저자명  
Zandehshahvar, Mohammadreza.
서명/저자  
Machine Learning for Knowledge Discovery in Engineering and Science: From Nanophotonic Design to Medical Diagnosis
발행사항  
[Sl] : Georgia Institute of Technology, 2023
발행사항  
Ann Arbor : ProQuest Dissertations & Theses, 2023
형태사항  
139 p
주기사항  
Source: Dissertations Abstracts International, Volume: 87-05, Section: B.
주기사항  
Advisor: Adibi, Ali.
학위논문주기  
Thesis (Ph.D.)--Georgia Institute of Technology, 2023.
초록/해제  
요약Artificial Intelligence (AI), particularly Machine Learning (ML) and Deep Learning (DL) tools, has made a transformative impact on various aspects of our lives. Their power to handle high-dimensional data has enabled them to solve problems across multiple scientific and engineering disciplines, from engineering new materials and designing novel nanostructures and sensors to analyzing unprecedented medical data. Despite these advancements, significant challenges remain in maximizing the capabilities of ML algorithms for specific scientific or engineering problems.A prevalent concern is that the ML algorithms, primarily designed for other tasks like image processing or pattern recognition, are often used directly in most studies without any significant tailoring to the problem of interest. Another under-explored aspect is knowledge discovery using ML to understand new phenomena in science and engineering by studying the influence of different parameters on a given outcome.This research aims to systematically address these concerns by proposing a universal approach that can be tailored and optimally adjusted to solve specific problems in different fields. This approach spans diverse problems, from the inverse design of nanostructures and nanosensors to diagnosis of lung diseas in radiology.Two high-impact applications are specifically selected to demonstrate the adaptability and efficacy of the proposed ML solutions: 1) optimization and knowledge discovery of photonic nanostructures, and 2) diagnosis and prognosis of of COVID-19 pneumonia from Chest X-ray (CXR).n the field of nanophotonics, the research introduced a novel approach based on Dimensionality Reduction (DR) of the design and response space, resulting in a considerable reduction in computational costs and enabling more efficient inverse design processes. The study of manifold learning and DR techniques for examining the feasibility range of responses in a class of nanostructures marks a significant breakthrough in the field. Additionally, a method for defining physics-friendly similarity measures tailored for nanophotonic design is developed based on deep metric learning.In radiology, significant strides are made in the ML-assisted diagnosis, focusing on labeling variability in CXRs, disease severity assessment, and the interpretability of DL models. A user-friendly online labeling tool is developed to gather data, study variability in labeling, and integrated human and machine assessments. A novel method using Bayesian Neural Networks (BNNs) is proposed for disease severity assessment. To enhance trust in AI, a unique visualization approach using a pruning method is also introduced.The dichotomy between these two applications underscores the adaptability and efficacy of the proposed ML tools in addressing significant challenges in different science and engineering fields. This research demonstrates the potential of tailored ML solutions in diverse domains, further proving their immense potential in resolving major challenges across a wide range of disciplines.
일반주제명  
Design optimization
일반주제명  
Pneumonia
일반주제명  
Knowledge discovery
일반주제명  
Neural networks
일반주제명  
Labeling
일반주제명  
Support vector machines
일반주제명  
Crystallization
일반주제명  
Thin films
일반주제명  
Disease transmission
일반주제명  
COVID-19
일반주제명  
Computer science
일반주제명  
Condensed matter physics
일반주제명  
Materials science
기타저자  
Georgia Institute of Technology.
기본자료저록  
Dissertations Abstracts International. 87-05B.
전자적 위치 및 접속  
로그인 후 원문을 볼 수 있습니다.

MARC

 008260203s2023        us                              c    eng  d
■001000017365980
■00520260209102907
■006m          o    d                
■007cr#unu||||||||
■020    ▼a9798263396695
■035    ▼a(MiAaPQ)AAI32315753
■035    ▼a(MiAaPQ)GeorgiaTech76742
■040    ▼aMiAaPQ▼cMiAaPQ
■0820  ▼a790
■1001  ▼aZandehshahvar,  Mohammadreza.
■24510▼aMachine  Learning  for  Knowledge  Discovery  in  Engineering  and  Science:  From  Nanophotonic  Design  to  Medical  Diagnosis
■260    ▼a[Sl]▼bGeorgia  Institute  of  Technology▼c2023
■260  1▼aAnn  Arbor▼bProQuest  Dissertations  &  Theses▼c2023
■300    ▼a139  p
■500    ▼aSource:  Dissertations  Abstracts  International,  Volume:  87-05,  Section:  B.
■500    ▼aAdvisor:  Adibi,  Ali.
■5021  ▼aThesis  (Ph.D.)--Georgia  Institute  of  Technology,  2023.
■520    ▼aArtificial  Intelligence  (AI),  particularly  Machine  Learning  (ML)  and  Deep  Learning  (DL)  tools,  has  made  a  transformative  impact  on  various  aspects  of  our  lives.  Their  power  to  handle  high-dimensional  data  has  enabled  them  to  solve  problems  across  multiple  scientific  and  engineering  disciplines,  from  engineering  new  materials  and  designing  novel  nanostructures  and  sensors  to  analyzing  unprecedented  medical  data.  Despite  these  advancements,  significant  challenges  remain  in  maximizing  the  capabilities  of  ML  algorithms  for  specific  scientific  or  engineering  problems.A  prevalent  concern  is  that  the  ML  algorithms,  primarily  designed  for  other  tasks  like  image  processing  or  pattern  recognition,  are  often  used  directly  in  most  studies  without  any  significant  tailoring  to  the  problem  of  interest.  Another  under-explored  aspect  is  knowledge  discovery  using  ML  to  understand  new  phenomena  in  science  and  engineering  by  studying  the  influence  of  different  parameters  on  a  given  outcome.This  research  aims  to  systematically  address  these  concerns  by  proposing  a  universal  approach  that  can  be  tailored  and  optimally  adjusted  to  solve  specific  problems  in  different  fields.  This  approach  spans  diverse  problems,  from  the  inverse  design  of  nanostructures  and  nanosensors  to  diagnosis  of  lung  diseas  in  radiology.Two  high-impact  applications  are  specifically  selected  to  demonstrate  the  adaptability  and  efficacy  of  the  proposed  ML  solutions:  1)  optimization  and  knowledge  discovery  of  photonic  nanostructures,  and  2)  diagnosis  and  prognosis  of  of  COVID-19  pneumonia  from  Chest  X-ray  (CXR).n  the  field  of  nanophotonics,  the  research  introduced  a  novel  approach  based  on  Dimensionality  Reduction  (DR)  of  the  design  and  response  space,  resulting  in  a  considerable  reduction  in  computational  costs  and  enabling  more  efficient  inverse  design  processes.  The  study  of  manifold  learning  and  DR  techniques  for  examining  the  feasibility  range  of  responses  in  a  class  of  nanostructures  marks  a  significant  breakthrough  in  the  field.  Additionally,  a  method  for  defining  physics-friendly  similarity  measures  tailored  for  nanophotonic  design  is  developed  based  on  deep  metric  learning.In  radiology,  significant  strides  are  made  in  the  ML-assisted  diagnosis,  focusing  on  labeling  variability  in  CXRs,  disease  severity  assessment,  and  the  interpretability  of  DL  models.  A  user-friendly  online  labeling  tool  is  developed  to  gather  data,  study  variability  in  labeling,  and  integrated  human  and  machine  assessments.  A  novel  method  using  Bayesian  Neural  Networks  (BNNs)  is  proposed  for  disease  severity  assessment.  To  enhance  trust  in  AI,  a  unique  visualization  approach  using  a  pruning  method  is  also  introduced.The  dichotomy  between  these  two  applications  underscores  the  adaptability  and  efficacy  of  the  proposed  ML  tools  in  addressing  significant  challenges  in  different  science  and  engineering  fields.  This  research  demonstrates  the  potential  of  tailored  ML  solutions  in  diverse  domains,  further  proving  their  immense  potential  in  resolving  major  challenges  across  a  wide  range  of  disciplines.
■590    ▼aSchool  code:  0078.
■650  4▼aDesign  optimization
■650  4▼aPneumonia
■650  4▼aKnowledge  discovery
■650  4▼aNeural  networks
■650  4▼aLabeling
■650  4▼aSupport  vector  machines
■650  4▼aCrystallization
■650  4▼aThin  films
■650  4▼aDisease  transmission
■650  4▼aCOVID-19
■650  4▼aComputer  science
■650  4▼aCondensed  matter  physics
■650  4▼aMaterials  science
■690    ▼a0800
■690    ▼a0984
■690    ▼a0611
■690    ▼a0794
■71020▼aGeorgia  Institute  of  Technology.
■7730  ▼tDissertations  Abstracts  International▼g87-05B.
■790    ▼a0078
■791    ▼aPh.D.
■792    ▼a2023
■793    ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17365980▼nKERIS▼z이  자료의  원문은  한국교육학술정보원에서  제공합니다.

Preview

Export

ChatGPT Discussion

AI Recommended Related Books


    New Books MORE
    Statistics for the past 3 years. Go to brief

    ค้นหาข้อมูลรายละเอียด

    • จองห้องพัก
    • ไม่อยู่
    • โฟลเดอร์ของฉัน
    • ขอดูแรก
    • Non-Book Loan Application
    • Nighttime Book Loan Application
    วัสดุ
    Reg No. Call No. ตำแหน่งที่ตั้ง สถานะ ยืมข้อมูล
    TF15941 전자도서 대출가능 My Folder 부재도서신고 비도서대출신청 야간 도서대출신청

    * จองมีอยู่ในหนังสือยืม เพื่อให้การสำรองที่นั่งคลิกที่ปุ่มจองห้องพัก

    Books borrowed together with this book

    Related Popular Books

    Available after logging in.