The ICSI/Berkeley Neural Theory of Language ProjectSlide 2Connectionist Model of Word Recognition (Rumelhart and McClelland)Constraints on Connectionist ModelsCan we formalize/model these intuitionsAbstract NeuronComputing with Abstract NeuronsDistributed vs Localist Rep’nSlide 9Slide 10Visual SystemCoarse CodingCoarse-Fine CodingConnectionist Models in Cognitive ScienceComputing other relationsTriangle nodes and McCullough-Pitts Neurons?“They all rose”Spreading activation and feature structuresRepresenting concepts using triangle nodesSlide 21Slide 22Categories and concepts- introductionLecture OutlineThe WCS Color ChipsConceptsSlide 29Concepts: Traditional TheoryThe neural theoryClassical vs prototype model of categorizationPrototype theorySlide 34Slide 36Slide 37Slide 38Categories - who decides?Basic-level categoriesCategories & Prototypes: OverviewBasic-level -- CriteriaSlide 44Slide 45Slide 46Basic-Level CategoryOther Basic-level categoriesConcepts are not categoricalMotherRadial Structure of MotherMarriageConcepts and radial categoriesCategory StructurePrototypeNeural Evidence for category structureCategory Naming and DeficitsA PET Study on categories (Nature 1996)StudyThe experimentSlide 61Slide 62Slide 63Slide 64Slide 65ConclusionsThe ICSI/Berkeley Neural Theory of Language ProjectLearning early constructions (Chang, Mok)ECGConnectionist Model of Word Recognition (Rumelhart and McClelland)Constraints on Connectionist Models100 Step Rule Human reaction times ~ 100 milliseconds Neural signaling time ~ 1 millisecondSimple messages between neuronsLong connections are rareNo new connections during learningDevelopmentally plausibleCan we formalize/model these intuitionsWhat is a neurally plausible computational model of spreading activation that captures these features.What does semantics mean in neurally embodied termsWhat are the neural substrates of concepts that underlie verbs, nouns, spatial predicates?Abstract Neuronw2wnw1w0I0 = 1o u t p u t yi2ini1. . .i n p u t iniiiiwnet0y1 if net > 00 otherwise{Computing with Abstract NeuronsMcCollough-Pitts Neurons were initially used to modelpattern classificationsize = small AND shape = round AND color = green AND location = on_tree => unripelinking classified patterns to behaviorsize = large OR motion = approaching => move_awaysize = small AND direction = above => move_aboveMcCollough-Pitts Neurons can compute logical functions. AND, NOT, ORDistributed vs Localist Rep’nJohn1 1 0 0Paul0 1 1 0George0 0 1 1Ringo1 0 0 1John1 0 0 0Paul0 1 0 0George0 0 1 0Ringo0 0 0 1What are the drawbacks of each representation?Distributed vs Localist Rep’nWhat happens if you want to represent a group?How many persons can you represent with n bits? 2^nWhat happens if one neuron dies?How many persons can you represent with n bits? nJohn1 1 0 0Paul0 1 1 0George0 0 1 1Ringo1 0 0 1John1 0 0 0Paul0 1 0 0George0 0 1 0Ringo0 0 0 1Sparse Distributed RepresentationVisual System1000 x 1000 visual mapFor each location, encode:orientationdirection of motionspeedsizecolordepthBlows up combinatorically!……Coarse Codinginfo you can encode with one fine resolution unit = info you can encode with a few coarse resolution unitsNow as long as we need fewer coarse units total, we’re goodCoarse-Fine Codingbut we can run into ghost “images”Feature 2e.g. Direction of MotionFeature 1e.g. OrientationYXGGY-OrientationX-OrientationY-Dir X-DirCoarse in F2, Fine in F1Coarse in F1, Fine in F2Connectionist Models in Cognitive ScienceStructured PDPNeural Conceptual Existence Data FittingHybridComputing other relationsThe 2/3 node is a useful function that activates its outputs (3) if any (2) of its 3 inputs are activeSuch a node is also called a triangle node and will be useful for lots of representations.Triangle nodes and McCullough-Pitts Neurons?B CAA B C“They all rose”triangle nodes:when two of the neurons fire, the third also firesmodel of spreading activationSpreading activation and feature structuresParallel activation streams.Top down and bottom up activation combine to determine the best matching structure.Triangle nodes bind features of objects to valuesMutual inhibition and competition between structuresMental connections are active neural connectionsRepresenting concepts using triangle nodesBarrett Ham Container Push dept~CS Color ~pink Inside ~region Schema ~slide sid~001 Taste ~salty Outside ~region Posture ~palm emp~GSI Bdy. ~curve Dir. ~ awaySchneider Pea Purchase Stroll dept~Ling Color ~green Buyer ~person Schema ~walk sid~002 Taste ~sweet Seller ~person Speed ~slow emp~Gra Cost ~money Dir. ~ ANY Goods ~ thingFeature Structures in Four DomainsCategories and concepts- introductionCS182/Ling109/CogSci110Spring 2008Lecture OutlineCategoriesBasic LevelPrototype EffectsNeural Evidence for Category StructureAspects of a Neural Theory of conceptsImage SchemasDescription and typesBehavioral Experiment on Image SchemasEvent Structure and Motor SchemasThe WCS Color ChipsBasic color terms:Single word (not blue-green)Frequently used (not mauve)Refers primarily to colors (not lime)Applies to any object (not blonde)ConceptsWhat Concepts Are: Basic ConstraintsConcepts are the elements of reason, and constitute the meanings of words and linguistic expressions.Concepts Are:•Universal: they characterize all particular instances; e.g., the concept of grasping is the same no matter who the agent is or what the patient is or how it is done.•Stable. •Internally structured. •Compositional.•Inferential. They interact to give rise to inferences.•Relational. They may be related by hyponymy, antonymy, etc.•Meaningful. •Not tied to the specific word forms used to express them.Concepts: Traditional TheoryThe Traditional TheoryReason and language are what distinguish human beings from other animals.Concepts therefore use only human-specific brain mechanisms.Reason is separate from perception and action, and does not make direct use of the sensory-motor system.Concepts must be “disembodied” in this sense.The neural theoryHuman concepts are embodied. Many concepts make direct use of sensory-motor, emotional, and social cognition capacities of our body-brain system.Many of these capacities are also present in non-human primates.Continuity Principle of Am.
View Full Document