본문

서브메뉴

Neural Software Abstractions- [electronic resource]
Neural Software Abstractions - [electronic resource]
Neural Software Abstractions- [electronic resource]

상세정보

자료유형  
 학위논문파일 국외
최종처리일시  
20240214100455
ISBN  
9798380382687
DDC  
004
저자명  
Chang, Michael.
서명/저자  
Neural Software Abstractions - [electronic resource]
발행사항  
[S.l.]: : University of California, Berkeley., 2023
발행사항  
Ann Arbor : : ProQuest Dissertations & Theses,, 2023
형태사항  
1 online resource(222 p.)
주기사항  
Source: Dissertations Abstracts International, Volume: 85-03, Section: B.
주기사항  
Advisor: Levine, Sergey;Griffiths, Thomas L.
학위논문주기  
Thesis (Ph.D.)--University of California, Berkeley, 2023.
사용제한주기  
This item must not be sold to any third party vendors.
초록/해제  
요약The desire to efficiently solve problems has driven humans to create tools to accomplish more with less. To be useful in a variety of contexts, a tool must encode knowledge of how to solve a general problem, knowledge that models the system that the tool manipulates. For most of human history, tools enabled humans to better manipulate only physical systems, such as using a lever for lifting heavy objects. These tools implicitly modeled the physical system via their specialized design. The computer is significant because it was the first universal tool for modeling and manipulating any system.Unfortunately, this universality has historically been restricted to systems that only humans can manually model and manipulate, via code. Humans have long acted as the interface between computers and the physical world, but we will increasingly become the bottleneck to progress as computers become more powerful and the world becomes more complex. If we could build machines that automatically model and manipulate systems on their own, then we would solve more problems with less effort: we would need only specify what the problem is rather than bother with how to solve it.The problem of building machines that automatically model and manipulate systems is not new and arguably encompasses the entire field of artificial intelligence (AI). Solving such a problem implies two things: first, that the machine can represent system interactions and second, that the machine can learn such representations automatically. What it means to represent system interactions is to represent the entities in the environment, the transformations that change the state of these entities, and choices the agent makes to apply these transformations. What it means to learn representations automatically is for these representations to be learned functions of the machine's raw sensorimotor stream. For such representations to be effective for automatically modeling and manipulating systems, they need to generalize over the combinatorial space of possible combinations of entities, of transformations, and of choices, and criterion that I call combinatorial generalization.Neither of the two paradigms that have dominated AI since the mid-1900s have yet offered a complete solution to both desiderata. The symbolic paradigm offers solutions for how to represent system interactions but not for how to learn representations. Conversely, the connectionist paradigm offers solutions for how to learn representations, but generally such representations do not directly expose the entities, transformations, and choices of the underlying system interaction in question. In the last half century these two paradigms have grown into the modern disciplines of software programming and deep learning, largely retaining their original complementary strengths and weaknesses. How can we achieve the strengths of both?One prominent class of approaches for combining both paradigms is to use neural networks for processing symbolic data or searching over symbolic code. These methods have achieved great success in natural language processing, code generation, and symbolic search, but they all assume a human-defined abstraction of the system to begin with.To actually address the problem of automatically modeling and manipulating systems, we need the machine to create these abstractions from its own sensorimotor experience. We need to combine both paradigms in a different way. What we would want instead are AI methods that can learn directly from raw data as deep learning algorithms do, with learned representations that generalize over the combinatorial space of system interactions as software does.My central thesis is that there is a deep similarity between electronic circuits and neural networks, and that adapting the methods we invented almost a century ago for creating modular software programs on top of analog circuits can enable neural networks to exhibit similar generalization properties as software does. I argue that the principle of separation of concerns was the key design principle that enabled representations in software to generalize and that contextual refinement was the key technique that enabled us to implement the principle of separation of concerns at every level of the computing stack. This thesis presents various ways for how to instantiate contextual refinement in neural networks and shows the gains in combinatorial generalization that this technique brings.
일반주제명  
Computer science.
일반주제명  
Computer engineering.
키워드  
Abstraction
키워드  
Compositionality
키워드  
Deep learning
키워드  
Generalization
키워드  
Machine learning
키워드  
Reinforcement learning
기타저자  
University of California, Berkeley Computer Science
기본자료저록  
Dissertations Abstracts International. 85-03B.
기본자료저록  
Dissertation Abstract International
전자적 위치 및 접속  
로그인 후 원문을 볼 수 있습니다.

MARC

 008240612s2023      us  |||||||||||||||c||eng  d
■001000016932416
■00520240214100455
■006m          o    d                
■007cr#unu||||||||
■020    ▼a9798380382687
■035    ▼a(MiAaPQ)AAI30492351
■040    ▼aMiAaPQ▼cMiAaPQ
■0820  ▼a004
■1001  ▼aChang,  Michael.
■24510▼aNeural  Software  Abstractions▼h[electronic  resource]
■260    ▼a[S.l.]:▼bUniversity  of  California,  Berkeley.  ▼c2023
■260  1▼aAnn  Arbor  :▼bProQuest  Dissertations  &  Theses,  ▼c2023
■300    ▼a1  online  resource(222  p.)
■500    ▼aSource:  Dissertations  Abstracts  International,  Volume:  85-03,  Section:  B.
■500    ▼aAdvisor:  Levine,  Sergey;Griffiths,  Thomas  L.
■5021  ▼aThesis  (Ph.D.)--University  of  California,  Berkeley,  2023.
■506    ▼aThis  item  must  not  be  sold  to  any  third  party  vendors.
■520    ▼aThe  desire  to  efficiently  solve  problems  has  driven  humans  to  create  tools  to  accomplish  more  with  less.  To  be  useful  in  a  variety  of  contexts,  a  tool  must  encode  knowledge  of  how  to  solve  a  general  problem,  knowledge  that  models  the  system  that  the  tool  manipulates.  For  most  of  human  history,  tools  enabled  humans  to  better  manipulate  only  physical  systems,  such  as  using  a  lever  for  lifting  heavy  objects.  These  tools  implicitly  modeled  the  physical  system  via  their  specialized  design.  The  computer  is  significant  because  it  was  the  first  universal  tool  for  modeling  and  manipulating  any  system.Unfortunately,  this  universality  has  historically  been  restricted  to  systems  that  only  humans  can  manually  model  and  manipulate,  via  code.  Humans  have  long  acted  as  the  interface  between  computers  and  the  physical  world,  but  we  will  increasingly  become  the  bottleneck  to  progress  as  computers  become  more  powerful  and  the  world  becomes  more  complex.  If  we  could  build  machines  that  automatically  model  and  manipulate  systems  on  their  own,  then  we  would  solve  more  problems  with  less  effort:  we  would  need  only  specify  what  the  problem  is  rather  than  bother  with  how  to  solve  it.The  problem  of  building  machines  that  automatically  model  and  manipulate  systems  is  not  new  and  arguably  encompasses  the  entire  field  of  artificial  intelligence  (AI).  Solving  such  a  problem  implies  two  things:  first,  that  the  machine  can  represent  system  interactions  and  second,  that  the  machine  can  learn  such  representations  automatically.  What  it  means  to  represent  system  interactions  is  to  represent  the  entities  in  the  environment,  the  transformations  that  change  the  state  of  these  entities,  and  choices  the  agent  makes  to  apply  these  transformations.  What  it  means  to  learn  representations  automatically  is  for  these  representations  to  be  learned  functions  of  the  machine's  raw  sensorimotor  stream.  For  such  representations  to  be  effective  for  automatically  modeling  and  manipulating  systems,  they  need  to  generalize  over  the  combinatorial  space  of  possible  combinations  of  entities,  of  transformations,  and  of  choices,  and  criterion  that  I  call  combinatorial  generalization.Neither  of  the  two  paradigms  that  have  dominated  AI  since  the  mid-1900s  have  yet  offered  a  complete  solution  to  both  desiderata.  The  symbolic  paradigm  offers  solutions  for  how  to  represent  system  interactions  but  not  for  how  to  learn  representations.  Conversely,  the  connectionist  paradigm  offers  solutions  for  how  to  learn  representations,  but  generally  such  representations  do  not  directly  expose  the  entities,  transformations,  and  choices  of  the  underlying  system  interaction  in  question.  In  the  last  half  century  these  two  paradigms  have  grown  into  the  modern  disciplines  of  software  programming  and  deep  learning,  largely  retaining  their  original  complementary  strengths  and  weaknesses.  How  can  we  achieve  the  strengths  of  both?One  prominent  class  of  approaches  for  combining  both  paradigms  is  to  use  neural  networks  for  processing  symbolic  data  or  searching  over  symbolic  code.  These  methods  have  achieved  great  success  in  natural  language  processing,  code  generation,  and  symbolic  search,  but  they  all  assume  a  human-defined  abstraction  of  the  system  to  begin  with.To  actually  address  the  problem  of  automatically  modeling  and  manipulating  systems,  we  need  the  machine  to  create  these  abstractions  from  its  own  sensorimotor  experience.  We  need  to  combine  both  paradigms  in  a  different  way.  What  we  would  want  instead  are  AI  methods  that  can  learn  directly  from  raw  data  as  deep  learning  algorithms  do,  with  learned  representations  that  generalize  over  the  combinatorial  space  of  system  interactions  as  software  does.My  central  thesis  is  that  there  is  a  deep  similarity  between  electronic  circuits  and  neural  networks,  and  that  adapting  the  methods  we  invented  almost  a  century  ago  for  creating  modular  software  programs  on  top  of  analog  circuits  can  enable  neural  networks  to  exhibit  similar  generalization  properties  as  software  does.  I  argue  that  the  principle  of  separation  of  concerns  was  the  key  design  principle  that  enabled  representations  in  software  to  generalize  and  that  contextual  refinement  was  the  key  technique  that  enabled  us  to  implement  the  principle  of  separation  of  concerns  at  every  level  of  the  computing  stack.  This  thesis  presents  various  ways  for  how  to  instantiate  contextual  refinement  in  neural  networks  and  shows  the  gains  in  combinatorial  generalization  that  this  technique  brings.
■590    ▼aSchool  code:  0028.
■650  4▼aComputer  science.
■650  4▼aComputer  engineering.
■653    ▼aAbstraction
■653    ▼aCompositionality
■653    ▼aDeep  learning
■653    ▼aGeneralization
■653    ▼aMachine  learning
■653    ▼aReinforcement  learning
■690    ▼a0984
■690    ▼a0800
■690    ▼a0464
■71020▼aUniversity  of  California,  Berkeley▼bComputer  Science.
■7730  ▼tDissertations  Abstracts  International▼g85-03B.
■773    ▼tDissertation  Abstract  International
■790    ▼a0028
■791    ▼aPh.D.
■792    ▼a2023
■793    ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T16932416▼nKERIS▼z이  자료의  원문은  한국교육학술정보원에서  제공합니다.
■980    ▼a202402▼f2024

미리보기

내보내기

chatGPT토론

Ai 추천 관련 도서


    신착도서 더보기
    최근 3년간 통계입니다.

    소장정보

    • 예약
    • 소재불명신고
    • 나의폴더
    • 우선정리요청
    • 비도서대출신청
    • 야간 도서대출신청
    소장자료
    등록번호 청구기호 소장처 대출가능여부 대출정보
    TF08874 전자도서 마이폴더 부재도서신고 비도서대출신청

    * 대출중인 자료에 한하여 예약이 가능합니다. 예약을 원하시면 예약버튼을 클릭하십시오.

    해당 도서를 다른 이용자가 함께 대출한 도서

    관련 인기도서

    로그인 후 이용 가능합니다.