본문

서브메뉴

Revisiting the Actor-Critic System in Striatum During Decision Making
Revisiting the Actor-Critic System in Striatum During Decision Making
Revisiting the Actor-Critic System in Striatum During Decision Making

상세정보

자료유형  
 학위논문 서양
최종처리일시  
20260202104840
ISBN  
9798297601062
DDC  
616
저자명  
Qu, Albert Jiaxu.
서명/저자  
Revisiting the Actor-Critic System in Striatum During Decision Making
발행사항  
[Sl] : University of California, Berkeley, 2025
발행사항  
Ann Arbor : ProQuest Dissertations & Theses, 2025
형태사항  
109 p
주기사항  
Source: Dissertations Abstracts International, Volume: 87-04, Section: B.
주기사항  
Advisor: Wilbrecht, Linda.
학위논문주기  
Thesis (Ph.D.)--University of California, Berkeley, 2025.
초록/해제  
요약The actor-critic framework has provided a foundational model of reinforcement learning in biological systems. In one of the most popular version of the model, the dorsal striatum is thought to function as the "actor" implementing action selection, while the ventral striatum functions as the "critic" for credit assignment. While this model is highly successful, experimental evidence increasingly suggests that the computational and neural mechanisms underlying decision making may contain additional nuance or even extend beyond this classic framework. In this thesis, I present experimental and computational work that serves to refine our understanding of both the critic and the actor, based on recordings made in the striatal circuits of mice during decision making.In the first study, I focused on the critic. Here, I investigated whether dopamine release in the nucleus accumbens reflects Bayesian inference computations beyond simple temporal difference learning. Using fiber photometry to measure dopamine signals during a probabilistic switching task, I compared predictions from Bayesian inference models against standard reinforcement learning models. Model comparison revealed that both mouse behavior and nucleus accumbens dopamine signals were better explained by models incorporating Bayesian inference for hidden state estimation, suggesting that the "critic" system performs sophisticated probabilistic computations rather than simple iterative value updates.In the second study, I focused on the actor. Here, I examined how the dorsal striatum implements active choice rejection, a process that challenges traditional models of action selection. Using bilateral recordings from direct and indirect pathway neurons during a serial decision-making task (Restaurant Row), I discovered that active rejection decisions involve distinct opponency patterns compared to active accepting decisions. Remarkably, rejection choices showed significant ipsilateral suppression, while acceptance choices showed classic contralateral activation. Optogenetic manipulations confirmed the causal role of this asymmetric activity pattern, with unilateral suppression of ipsilateral direct pathway neurons specifically enhancing rejection of marginally valuable offers. To conclude, I will discuss how these findings might be used to revise and update the actor-critic framework. I propose that the neural model of the "critic" component be readily expanded to include the possibility of Bayesian inference mechanisms that enable rapid adaptation to environmental volatility through belief state representations. Then, I suggest that the neural model of the "actor" system can also be improved by including hemispheric opponency, while also accounting for hemispheric specialization in cost and benefit processing, when lateralized choice is being implemented. Together, these results demonstrate that biological decision making involves computational sophistication and circuit complexity that extends well beyond traditional reinforcement learning models, providing new insights into the neural basis of adaptive behavior and offering refined frameworks for understanding decision-making disorders.
일반주제명  
Neurosciences
일반주제명  
Bioinformatics
일반주제명  
Computer science
키워드  
Actor critic
키워드  
Basal ganglia
키워드  
Bayesian inference
키워드  
Decision making
키워드  
Reinforcement learning
키워드  
Striatal circuits
기타저자  
University of California, Berkeley Psychology
기본자료저록  
Dissertations Abstracts International. 87-04B.
전자적 위치 및 접속  
로그인 후 원문을 볼 수 있습니다.

MARC

 008260126s2025        us                              c    eng  d
■001000017359141
■00520260202104840
■006m          o    d                
■007cr#unu||||||||
■020    ▼a9798297601062
■035    ▼a(MiAaPQ)AAI32172833
■040    ▼aMiAaPQ▼cMiAaPQ
■0820  ▼a616
■1001  ▼aQu,  Albert  Jiaxu.
■24510▼aRevisiting  the  Actor-Critic  System  in  Striatum  During  Decision  Making
■260    ▼a[Sl]▼bUniversity  of  California,  Berkeley▼c2025
■260  1▼aAnn  Arbor▼bProQuest  Dissertations  &  Theses▼c2025
■300    ▼a109  p
■500    ▼aSource:  Dissertations  Abstracts  International,  Volume:  87-04,  Section:  B.
■500    ▼aAdvisor:  Wilbrecht,  Linda.
■5021  ▼aThesis  (Ph.D.)--University  of  California,  Berkeley,  2025.
■520    ▼aThe  actor-critic  framework  has  provided  a  foundational  model  of  reinforcement  learning  in  biological  systems.  In  one  of  the  most  popular  version  of  the  model,  the  dorsal  striatum  is  thought  to  function  as  the  "actor"  implementing  action  selection,  while  the  ventral  striatum  functions  as  the  "critic"  for  credit  assignment.  While  this  model  is  highly  successful,  experimental  evidence  increasingly  suggests  that  the  computational  and  neural  mechanisms  underlying  decision  making  may  contain  additional  nuance  or  even  extend  beyond  this  classic  framework.  In  this  thesis,  I  present  experimental  and  computational  work  that  serves  to  refine  our  understanding  of  both  the  critic  and  the  actor,  based  on  recordings  made  in  the  striatal  circuits  of  mice  during  decision  making.In  the  first  study,  I  focused  on  the  critic.  Here,  I  investigated  whether  dopamine  release  in  the  nucleus  accumbens  reflects  Bayesian  inference  computations  beyond  simple  temporal  difference  learning.  Using  fiber  photometry  to  measure  dopamine  signals  during  a  probabilistic  switching  task,  I  compared  predictions  from  Bayesian  inference  models  against  standard  reinforcement  learning  models.  Model  comparison  revealed  that  both  mouse  behavior  and  nucleus  accumbens  dopamine  signals  were  better  explained  by  models  incorporating  Bayesian  inference  for  hidden  state  estimation,  suggesting  that  the  "critic"  system  performs  sophisticated  probabilistic  computations  rather  than  simple  iterative  value  updates.In  the  second  study,  I  focused  on  the  actor.  Here,  I  examined  how  the  dorsal  striatum  implements  active  choice  rejection,  a  process  that  challenges  traditional  models  of  action  selection.  Using  bilateral  recordings  from  direct  and  indirect  pathway  neurons  during  a  serial  decision-making  task  (Restaurant  Row),  I  discovered  that  active  rejection  decisions  involve  distinct  opponency  patterns  compared  to  active  accepting  decisions.  Remarkably,  rejection  choices  showed  significant  ipsilateral  suppression,  while  acceptance  choices  showed  classic  contralateral  activation.  Optogenetic  manipulations  confirmed  the  causal  role  of  this  asymmetric  activity  pattern,  with  unilateral  suppression  of  ipsilateral  direct  pathway  neurons  specifically  enhancing  rejection  of  marginally  valuable  offers. To  conclude,  I  will  discuss  how  these  findings  might  be  used  to  revise  and  update  the  actor-critic  framework.  I  propose  that  the  neural  model  of  the  "critic"  component  be  readily  expanded  to  include  the  possibility  of  Bayesian  inference  mechanisms  that  enable  rapid  adaptation  to  environmental  volatility  through  belief  state  representations.  Then,  I  suggest  that  the  neural  model  of  the  "actor"  system  can  also  be  improved  by  including  hemispheric  opponency,  while  also  accounting  for  hemispheric  specialization  in  cost  and  benefit  processing,  when  lateralized  choice  is  being  implemented.  Together,  these  results  demonstrate  that  biological  decision  making  involves  computational  sophistication  and  circuit  complexity  that  extends  well  beyond  traditional  reinforcement  learning  models,  providing  new  insights  into  the  neural  basis  of  adaptive  behavior  and  offering  refined  frameworks  for  understanding  decision-making  disorders.
■590    ▼aSchool  code:  0028.
■650  4▼aNeurosciences
■650  4▼aBioinformatics
■650  4▼aComputer  science
■653    ▼aActor  critic
■653    ▼aBasal  ganglia
■653    ▼aBayesian  inference
■653    ▼aDecision  making
■653    ▼aReinforcement  learning
■653    ▼aStriatal  circuits
■690    ▼a0317
■690    ▼a0715
■690    ▼a0984
■71020▼aUniversity  of  California,  Berkeley▼bPsychology.
■7730  ▼tDissertations  Abstracts  International▼g87-04B.
■790    ▼a0028
■791    ▼aPh.D.
■792    ▼a2025
■793    ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17359141▼nKERIS▼z이  자료의  원문은  한국교육학술정보원에서  제공합니다.

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
    TF17629 전자도서 대출가능 마이폴더 부재도서신고 비도서대출신청 야간 도서대출신청

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

    해당 도서를 다른 이용자가 함께 대출한 도서

    관련 인기도서

    로그인 후 이용 가능합니다.