神經(jīng)信息學(xué)Neuroinformatics課件

神經(jīng)信息學(xué)NeuroinformaticsSpring

semester,2009LECTURE1Introduction武志華·

中科院生物物理所，腦與認(rèn)知科學(xué)國(guó)家重點(diǎn)實(shí)驗(yàn)室，副研究員·

Tel:64869355

Email:wuzh@史忠植·

中科院計(jì)算技術(shù)研究所,智能科學(xué)實(shí)驗(yàn)室，研究員·

人工智能

（神經(jīng)計(jì)算）教學(xué)按排每周四:

8:50-11:4040學(xué)時(shí),2學(xué)分3學(xué)時(shí)/每次地點(diǎn)：玉泉路園區(qū)

305

閉卷筆試筆試內(nèi)容與作業(yè)的關(guān)系引言－

1994年,汪云九老師在中科院研究生院開(kāi)設(shè)“神經(jīng)信息學(xué)”課程。

－

WhatisNeuroinformatics?

WhatisI.Neuroinformatics?ComputationalNeuroscience(計(jì)算神經(jīng)科學(xué))?III.TheoreticalNeuroscience?Neurocomputing(神經(jīng)計(jì)算)?

Whywelearn“Neuroinformatics”or“ComputationalNeuroscience”?VI.Coursestructure1.databasesofneurosciencedata,2.toolsformanagement,sharing,

analyzingandmodelingofneurosciencedata

atalllevelsofanalysis,3.computationalmodelsofthenervoussystemandneuralprocessesNeuroinformatics

isaresearchfieldincludingthedevelopmentof:Neuro-scienceInformationscienceNeuro-informaticsVastamountsofdiversedataaboutthebrainwasgathered(汪云九老師)HumanBrainProject1993年“人類腦計(jì)劃（HumanBrainProject）”的第一批項(xiàng)目公布，標(biāo)志著人類腦計(jì)劃正式啟動(dòng)Ideaemergedin1991:Mappingthebrainanditsfunctions.IntegratingenablingtechnologiesintoneuroscienceresearchNeuroinformaticsusesdatabases,theInternet,andvisualizationinthestorageandanalysisoftheneurosciencedata(AnamesimilartoHumanGenomeProject)

SenseLab(http://)132BrainM

(http://)

ExploreBrainMapsdatain3D

Neuroinformatics

=

Databases+Tools+ComputationalModelsWhatis

I.Neuroinformatics?Neurocomputing(神經(jīng)計(jì)算)?

III.TheoreticalNeuroscience?ComputationalNeuroscience(計(jì)算神經(jīng)科學(xué))?Whywelearn“Neuroinformatics”or“ComputationalNeuroscience”?VI.Coursestructure

Neurocomputingisconcernedwithprocessinginformation:1.Itinvolvesalearningprocesswithinanartificialneuralnetworkarchitecture2.Thetrainednetworkscanbeusedtoperformcertaintasksdependingontheparticularapplication3.Neurocomputingcanplayanimportantroleinsolvingcertaindifficultproblemsinscienceandengineeringsuchaspatternrecognition,optimization,eventclassification,controlandidentificationofnonlinearsystems,andstatisticalanalysisI.WhatisNeuroinformatics?Neurocomputing(神經(jīng)計(jì)算)?

III.TheoreticalNeuroscience?ComputationalNeuroscience(計(jì)算神經(jīng)科學(xué))?Whywelearn“Neuroinformatics”or“ComputationalNeuroscience”?

VI.CoursestructureTheoreticalNeuroscience=ComputationalandMathematicalModelingofNeuralSystems=Computational

NeuroscienceWhatisI.Neuroinformatics?Neurocomputing(神經(jīng)計(jì)算)?

III.TheoreticalNeuroscience?ComputationalNeuroscience(計(jì)算神經(jīng)科學(xué))?Whywelearn“Neuroinformatics”or“ComputationalNeuroscience”?

VI.CoursestructureComputationalneuroscienceisasubfieldofneurosciencethatusesmathematicalmethodstosimulateandunderstandthefunctionofthenervoussystem

(http://)Neuro-scienceArtificalNeuralnetworksDynamicalsystemComputationalneuroscienceComputationalneuroscienceAfamilymemberofbrainscience.ComputersimulationsofneuronsandneuralnetworksarecomplementarytotraditionaltechniquesinneuroscienceTheoreticalanalysisandcomputationalmodelingareimportanttoolsforcharacterizingwhatnervoussystemsdo,determininghowtheyfunction,andunderstandingwhytheyoperateinparticularwaysNeuroscienceencompassesapproachesrangingfrommolecularandcellularstudiestohumanpsychophysicsandpsychologyTheoreticalneuroscienceencouragescross-talkamongthesesubdisciplinesbyconstructingcompactrepresentationsofwhathasbeenlearned,buildingbridgesbetweendifferentlevelsofdescription,andidentifyingunifyingconceptsandprinciplesFirstneuronmodel－McCulloch&Pittsmodel(1943)(BulletinofMathematicalBiophysics5:115-133)Range:(0,1)or(-1,1)Time:t

t+1MPmodelsimulatesafewproperties?Theunithastwostatesdependingonthethreshold:restoractivated?Twotypesofsynapses:inhibitoryandexcitatory?Theunitreceivesthelinearsumofallthepre-synapticinputs

?Theintroductionoftime,mimickingthesynapticdelayAdvantage:Beabletoperformlogicoperations

Shortcoming:ToosimpletomodeltherealneuronGoalThefirstgoalistoteachWHYmathematicalandcomputationalmethods

areimportantinunderstandingthestructure,functionanddynamics

ofneuralorganization2.ThesecondgoalistoexplainHOWneural

phenomenaoccurringatdifferenthierarchicallevelscanbedescribed

bypropermathematicalmodelsWhatisI.Neuroinformatics?Neurocomputing(神經(jīng)計(jì)算)?

III.TheoreticalNeuroscience?ComputationalNeuroscience(計(jì)算神經(jīng)科學(xué))?Whywelearn“Neuroinformatics”or

“ComputationalNeuroscience”?

VI.CoursestructureWhynotgooutforawalk?

ImeanthecurrentneuroscienceworldisalittledifferentfrombeforeForexample:TheAgeofBrainScience（Japan）?UnderstandingtheBrain?

ProtectingtheBrain?

CreatingtheBrain?NurturingtheBrain（October1997---March2008）Ifyouareastudentmajoringinexperimentalbiology,－youmayfigureoutyourexperimentalproblemoranalyzeresultsinadifferentwayIfyouareastudenthavinggoodmathematicalorphysicalbackground,－???infutureAboutmathematicalandphysicalFORMULA:

－HighschoolknowledgeisenoughCaveatComputationalneuroscienceisahugeandfastdevelopingareaThisisonlyaveryshortcourse,andhopefullyitcanprovide:－Thebasicconceptsandmethodsofcomputationalneuroscienceresearch－Somebriefintroductiontoneurobiologicalconceptsandmathematicaltechniques.ThetechniqueswillbeappliedfordescribingthebehaviorofseveralbrainregionsWhatisI.Neuroinformatics?Neurocomputing(神經(jīng)計(jì)算)?

III.TheoreticalNeuroscience?ComputationalNeuroscience(計(jì)算神經(jīng)科學(xué))?Whywelearn“Neuroinformatics”or

“ComputationalNeuroscience”?

VI.CoursestructureCourseStructure0.

Introduction:Whatiscomputationalneuroscience,whyisiturgentlyneeded1.Singleneuronmodels2.

NeuralnetworkmodelsNeuralCodingSynapticplasticityandlearningHotpointsinbrainmodelingNeurocom