본문

서브메뉴

Building Features in Visual Neural Networks
Building Features in Visual Neural Networks
Building Features in Visual Neural Networks

Detailed Information

자료유형  
 학위논문 서양
최종처리일시  
20250211151433
ISBN  
9798382776965
DDC  
150
저자명  
Hamblin, Christopher.
서명/저자  
Building Features in Visual Neural Networks
발행사항  
[Sl] : Harvard University, 2024
발행사항  
Ann Arbor : ProQuest Dissertations & Theses, 2024
형태사항  
183 p
주기사항  
Source: Dissertations Abstracts International, Volume: 85-12, Section: B.
주기사항  
Advisor: Alvarez, George;Konkle, Talia.
학위논문주기  
Thesis (Ph.D.)--Harvard University, 2024.
초록/해제  
요약Deep neural networks have emerged as the dominant model class of the human visual system, however the algorithms they implement are often thought to be inscrutable. In this dissertation, we push back against this characterization, particularly as it applies to neural networks for object classification. Analogous to the ventral visual stream, these models are thought to compute a hierarchy of feature detectors, with features in each layer of the hierarchy computed as a function of those features represented in the previous layer. Features in early layers are thought to detect simple things, like colors and edges, while those in deep layers might detect complex constructs, like 'Junco feathers'. A complete understanding of object recognition requires that we reconcile not only what features are represented in this hierarchy, but how rich features are computed through the composition of simpler constituents. In three chapters, we will develop several interpretability tools for deep neural networks, aimed at providing visually intuitive explanations of feature construction.In Chapter 1, we will combat a major obstacle to our understanding of feature construction -- that features are embedded in large networks with many parameters. A consequence of this is that when one simply considers the networks architecture, the function that computes any feature in the network also has many parameters. We propose a technique for circuit pruning, which specifies a sparse route through the network by which a given feature computes its response to an input image(s). In Chapter 2, we propose a novel technique for visualizing what a feature responds to in a particular input image, which we call 'feature accentuation'. Typically, explanations of feature responses to individual images rely on attribution maps, which are displayed as heatmaps over the image highlighting the most exciting regions. However, an explanation of where important regions are located is insufficient; what is it the feature sees in these locations? With feature accentuation, we exaggerate the expression of a feature in a given input with gradient-based activation maximization, revealing how even when two features are excited by the same region of an image, it is often for very different reasons. In Chapter 3, we will address the functional role of feature inhibition; that is, what are the mechanisms by which the model ensures images do not express a given feature? Inhibition has received far less treatment in the literature than excitation, yet is critical for the construction of discriminative features. We observe that standard interpretability tools are not immediately suited to the inhibitory case, given the asymmetry introduced by the ReLU activation function. Given this, we propose inhibition be understood through a study of maximally tense images, i.e. those images that excite and inhibit a given feature simultaneously.
일반주제명  
Psychology
일반주제명  
Computer engineering
키워드  
Convolutional neural networks
키워드  
Explainable AI
키워드  
Feature visualization
키워드  
Mechanistic interpretability
키워드  
Neural circuits
키워드  
Neural network Interpretability
기타저자  
Harvard University Psychology
기본자료저록  
Dissertations Abstracts International. 85-12B.
전자적 위치 및 접속  
로그인 후 원문을 볼 수 있습니다.

MARC

 008250123s2024        us                              c    eng  d
■001000017161701
■00520250211151433
■006m          o    d                
■007cr#unu||||||||
■020    ▼a9798382776965
■035    ▼a(MiAaPQ)AAI31295428
■040    ▼aMiAaPQ▼cMiAaPQ
■0820  ▼a150
■1001  ▼aHamblin,  Christopher.▼0(orcid)0009-0003-0723-5961
■24510▼aBuilding  Features  in  Visual  Neural  Networks
■260    ▼a[Sl]▼bHarvard  University▼c2024
■260  1▼aAnn  Arbor▼bProQuest  Dissertations  &  Theses▼c2024
■300    ▼a183  p
■500    ▼aSource:  Dissertations  Abstracts  International,  Volume:  85-12,  Section:  B.
■500    ▼aAdvisor:  Alvarez,  George;Konkle,  Talia.
■5021  ▼aThesis  (Ph.D.)--Harvard  University,  2024.
■520    ▼aDeep  neural  networks  have  emerged  as  the  dominant  model  class  of  the  human  visual  system,  however  the  algorithms  they  implement  are  often  thought  to  be  inscrutable.  In  this  dissertation,  we  push  back  against  this  characterization,  particularly  as  it  applies  to  neural  networks  for  object  classification.  Analogous  to  the  ventral  visual  stream,  these  models  are  thought  to  compute  a  hierarchy  of  feature  detectors,  with  features  in  each  layer  of  the  hierarchy  computed  as  a  function  of  those  features  represented  in  the  previous  layer.  Features  in  early  layers  are  thought  to  detect  simple  things,  like  colors  and  edges,  while  those  in  deep  layers  might  detect  complex  constructs,  like  'Junco  feathers'.  A  complete  understanding  of  object  recognition  requires  that  we  reconcile  not  only  what  features  are  represented  in  this  hierarchy,  but  how  rich  features  are  computed  through  the  composition  of  simpler  constituents.  In  three  chapters,  we  will  develop  several  interpretability  tools  for  deep  neural  networks,  aimed  at  providing  visually  intuitive  explanations  of  feature  construction.In  Chapter  1,  we  will  combat  a  major  obstacle  to  our  understanding  of  feature  construction  --  that  features  are  embedded  in  large  networks  with  many  parameters.  A  consequence  of  this  is  that  when  one  simply  considers  the  networks  architecture,  the  function  that  computes  any  feature  in  the  network  also  has  many  parameters.  We  propose  a  technique  for  circuit  pruning,  which  specifies  a  sparse  route  through  the  network  by  which  a  given  feature  computes  its  response  to  an  input  image(s).  In  Chapter  2,  we  propose  a  novel  technique  for  visualizing  what  a  feature  responds  to  in  a  particular  input  image,  which  we  call  'feature  accentuation'.  Typically,  explanations  of  feature  responses  to  individual  images  rely  on  attribution  maps,  which  are  displayed  as  heatmaps  over  the  image  highlighting  the  most  exciting  regions.  However,  an  explanation  of  where  important  regions  are  located  is  insufficient;  what  is  it  the  feature  sees  in  these  locations?  With  feature  accentuation,  we  exaggerate  the  expression  of  a  feature  in  a  given  input  with  gradient-based  activation  maximization,  revealing  how  even  when  two  features  are  excited  by  the  same  region  of  an  image,  it  is  often  for  very  different  reasons.  In  Chapter  3,  we  will  address  the  functional  role  of  feature  inhibition;  that  is,  what  are  the  mechanisms  by  which  the  model  ensures  images  do  not  express  a  given  feature?  Inhibition  has  received  far  less  treatment  in  the  literature  than  excitation,  yet  is  critical  for  the  construction  of  discriminative  features.  We  observe  that  standard  interpretability  tools  are  not  immediately  suited  to  the  inhibitory  case,  given  the  asymmetry  introduced  by  the  ReLU  activation  function.  Given  this,  we  propose  inhibition  be  understood  through  a  study  of  maximally  tense  images,  i.e.  those  images  that  excite  and  inhibit  a  given  feature  simultaneously.
■590    ▼aSchool  code:  0084.
■650  4▼aPsychology
■650  4▼aComputer  engineering
■653    ▼aConvolutional  neural  networks
■653    ▼aExplainable  AI
■653    ▼aFeature  visualization
■653    ▼aMechanistic  interpretability
■653    ▼aNeural  circuits
■653    ▼aNeural  network  Interpretability
■690    ▼a0800
■690    ▼a0621
■690    ▼a0464
■71020▼aHarvard  University▼bPsychology.
■7730  ▼tDissertations  Abstracts  International▼g85-12B.
■790    ▼a0084
■791    ▼aPh.D.
■792    ▼a2024
■793    ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17161701▼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

    Подробнее информация.

    • Бронирование
    • не существует
    • моя папка
    • Первый запрос зрения
    • Non-Book Loan Application
    • Nighttime Book Loan Application
    материал
    Reg No. Количество платежных Местоположение статус Ленд информации
    TF11332 전자도서 대출가능 My Folder 부재도서신고 비도서대출신청 야간 도서대출신청

    * Бронирование доступны в заимствований книги. Чтобы сделать предварительный заказ, пожалуйста, нажмите кнопку бронирование

    Books borrowed together with this book

    Related Popular Books

    Available after logging in.