Connectionism and models of memory and amnesia Jaap Murre University of Amsterdam

Slides:



Advertisements
Similar presentations
N Waarvoor wordt Myke Tyson geschorst in de wedstrijd tegen Evander Holyfield bij de wereldtitel zwaargewicht boksen? © Amsterdamse Media Vragenlijst.
Advertisements

Remembering & Forgetting
Chapter 7: Human Memory. Human Memory: Basic Questions  How does information get into memory?  How is information maintained in memory?  How is information.
Memory and Hippocampus By:Mohammad Ali Ahmadi-Pajouh All Materials are from “Principals of Neuroscience” Written by E. Kandel In the Name of Allah Amirkabir.
Long-Term Memory: Encoding and Retrieval
Neuropsychology of amnesia
Lecture 3: Learning and Memory Prof.dr. Jaap Murre University of Maastricht University of Amsterdam
Lecture Overview The Nature of Memory Forgetting Biological Bases of Memory ©John Wiley & Sons, Inc
Learning and Memory in Hippocampus and Neocortex: A Complementary Learning Systems Approach Psychology 209 Feb 11, 2014.
Learning in Recurrent Networks Psychology 209 February 25, 2013.
Section 7 Learning and Memory. I Learning Learning: associative and nonassociative The acquisition of knowledge or skill; Associate and nonassociative.
COGNITIVE NEUROSCIENCE
Mind, Brain & Behavior Friday March 14, What to Study for the Final Exam  Chapters 26 & 28 – Motor Activity Know what kind of info the two main.
Chapter Seven The Network Approach: Mind as a Web.
Long Term Memory Long Term Memory Scott Betournay.
Biologically Inspired Robotics Group,EPFL Associative memory using coupled non-linear oscillators Semester project Final Presentation Vlad TRIFA.
Memory Systems Chapter 23 Friday, December 5, 2003.
Thanks for the memories Functional aspects of memory Richard Fielding Department of Community Medicine HKU.
Memory Consolidation A Summary PSY 506A Molly Bisbee.
Cooperation of Complementary Learning Systems in Memory Review and Update on the Complementary Learning Systems Framework James L. McClelland Psychology.
COGNITIVE SCIENCE 17 Can You Remember My Name? Part 1 Jaime A. Pineda, Ph.D.
Memory and Consolidation Prof.dr. Jaap Murre University of Amsterdam University of Maastricht
Learning, memory & amnesia
Memory. Encoding, Retrieval, and Recall Types of Memory (Explicit)(Implicit)
Brain Plasticity and the Stability of Cognition Studies in Cognitive Neuroscience Jaap Murre University of Amsterdam.
Module 12 Remembering & Forgetting. INTRODUCTION recall –retrieving previously learned information without the aid of or with very few external cues recognition.
4 th Edition Copyright 2004 Prentice Hall7-1 Memory Chapter 7.
Chapter 8: Human Memory. Human Memory: Basic Questions How does information get into memory? How is information maintained in memory? How is information.
Chapter 7 Human Memory. Table of Contents Human Memory: Basic Questions How does information get into memory? How is information maintained in memory?
© 2000 John Wiley & Sons, Inc. Huffman/Vernoy/Vernoy: Psychology in Action 5e Psychology in Action, Fifth Edition by Karen Huffman, Mark Vernoy, and Judith.
Memory Do we remember from stories our parents tell us or are they genuine? Why can I remember every detail of what and where I was when I found out John.
Chapter 6 Memory.
The Brain Basis of Memory: Theory and Data James L. McClelland Stanford University.
 How does memory affect your identity?  If you didn’t have a memory how would your answer the question – How are you today?
Companion website: MEMORY.
MULTIPLE MEMORY SYSTEM IN HUMANS
Memory The brain’s system for filing away new information and retrieving previously learned data A constructive process 3 types of memory Sensory memory.
Relational Learning and Amnesia
Retrieval & Retrieval Failure.  What is the serial position effect?  What are flashbulb memories?  What is the forgetting curve?  What is the difference.
HUH? : WHEN MEMORY LAPSES.  Hermann Ebbinghaus tested memory  Created Forgetting Curve: graphs retention and forgetting over time  Showed steep drop.
What causes Forgetting ? Biological or organic causes are the basis for a lot of forgetting. This Usually refers to damage to the brain brought about by:
Module 12 Remembering & Forgetting. INTRODUCTION Recall –Retrieving previously learned information without the aid of, or with very few, external cues.
Chapter 7 Memory. What is MEMORY? Memory – internal record of some prior event or experience; a set of mental processes that receives, encodes, stores,
Chapter 7: Human Memory.
Neural Networks Presented by M. Abbasi Course lecturer: Dr.Tohidkhah.
Memory How do we retain information? How do we recall information?
Memory Li, Kristoffer Daniel Lee, Seoui. What is Memory? An active system that receives information from the senses, puts that information into usable.
Encoding StorageRetrievalForgetting Research and People.
MEMORY PROF ELHAM Aljammas May 2015 L16 © 2002 John Wiley & Sons, Inc. Huffman: PSYCHOLOGY IN ACTION, 6E.
Chapter 6 Memory. The mental processes that enable us to retain and sue information over time.
Memory: Its Nature and Organization in the Brain James L. McClelland Stanford University.
Ch 11: Learning, Memory & Amnesia
Chapter 7 Notes AP Tips. Be able to identify to three steps necessary to have memories. Encoding: the process of acquiring and entering information into.
The Neuropsychology of Memory Ch. 11. Outline Case studies Korsakoff’s Amnesia Alzheimer’s Disease Posttraumatic Amnesia Clive Wearing video Theories.
Psychology in Action (8e) by Karen Huffman
Long-term Memory Explicit Memories (fact-based info, conscious retrieval) Semantic memories (memory of facts) Episodic memories (events) Implicit Memories.
Introduction to Connectionism Jaap Murre Universiteit van Amsterdam en Universiteit Utrecht
Memory: An Introduction
What is cognitive psychology?
Psychology 209 – Winter 2017 Feb 28, 2017
Neural Networks.
Cooperation of Complementary Learning Systems in Memory
Psychology in Action (8e) by Karen Huffman
Cognitive Processes PSY 334
Memory Gateway to Learning.
Remembering & Forgetting
CLS, Rapid Schema Consistent Learning, and Similarity-weighted Interleaved learning Psychology 209 Feb 26, 2019.
The Network Approach: Mind as a Web
Remembering & Forgetting
thinking about learning and memory
Presentation transcript:

Connectionism and models of memory and amnesia Jaap Murre University of Amsterdam

The French neurologist Ribot discovered more than 100 years ago that in retrograde amnesia one tends to loose recent memories Memory loss gradients in RA are called Ribot gradients

Overview Catastrophic interference and hypertransfer Brief review of neuroanatomy Outline of the TraceLink model Some simulation results of neural network model, focussing on retrograde amnesia Recent work: –Mathematical point-process model Concluding remarks

Catastrophic interference Learning new patterns in backpropation will overwrite all existing patterns Rehearsal is necessary McCloskey and Cohen (1989), Ratcliff (1990) This is not psychologically plausible

Osgood surface (1949) Paired-associates in lists A and B will interfere strongly if the stimuli are similar but the responses vary If stimuli are different, little interference (i.e., forgetting) occurs Backpropagation also shows odd behavior if stimuli vary but responses are similar in lists A and B (hypertransfer)

Learned responses StimuliTarget responses (after three learning trials) Phase 1: Learning list A rist munk twup gork gomp toup wemp twub twup Phase 2: Learning interfering list B (after five learning trials) yupe munk muup maws gomp twup drin twub twub Phase 3: Retesting on list A rist munk goub gork gomp tomp wemp twub twub Hypertransfer

Problems with sequential learning in backpropagation Reason 1: Strongly overlapping hidden- layer representations Remedy 1: reduce the hidden-layer representations –French, Murre (semi-distributed representations)

Problems with sequential learning in backpropagation Reason 2: Satisfying only immediate learning constraints Remedy 2: Rehearse some old patterns, when learning new ones –Murre (1992): random rehearsal –McClelland, McNaughton and O’Reilly (1995): interleaved learning

Final remarks on sequential learning Two-layer ‘backpropagation’ networks do show plausible forgetting Other learning networks do not exhibit catastrophic interference: ART, CALM, Kohonen Maps, etc. It is not a necessary condition of learning neural networks; it mainly affects backpropagation The brain does not do backpropagation and therefore does not suffer from this problem

Models of amnesia and memory in the brain TraceLink Point-process model Chain-development model

Neuroanatomy of amnesia Hippocampus Adjacent areas such as entorhinal cortex and parahippocampal cortex Basal forebrain nuclei Diencephalon

The position of the hippocampus in the brain

Hippocampal connections

Hippocampus has an excellent overview of the entire cortex

Trace-Link model: structure

System 1: Trace system Function: Substrate for bulk storage of memories, ‘association machine’ Corresponds roughly to neocortex

System 2: Link system Function: Initial ‘scaffold’ for episodes Corresponds roughly to hippocampus and certain temporal and perhaps frontal areas

System 3: Modulatory system Function: Control of plasticity Involves at least parts of the hippocampus, amygdala, fornix, and certain nuclei in the basal forebrain and in the brain stem

Stages in episodic learning

Dreaming and consolidation of memory Theory by Francis Crick and Graeme Mitchison (1983) Main problem: Overloading of memory Solution: Reverse learning leads to removal of ‘obsessions’ “We dream in order to forget”

Dreaming and memory consolidation When should this reverse learning take place? During REM sleep –Normal input is deactivated –Semi-random activations from the brain stem –REM sleep may have lively hallucinations

Consolidation may also strengthen memory This may occur during deep sleep (as opposed to REM sleep) Both hypothetical processes may work together to achieve an increase in the definition of representations in the cortex

Recent data by Matt Wilson and Bruce McNaughton (1994) 120 neurons in rat hippocampus PRE: Slow-wave sleep before being in the experimental environment (cage) RUN: During experimental environment POST: Slow-wave sleep after having been in the experimental environment

Wilson en McNaughton Data PRE: Slow-wave sleep before being in the experimental environment (cage) RUN: During experimental environment POST: Slow-wave sleep after having been in the experimental environment

Some important characteristics of amnesia Anterograde amnesia (AA) –Implicit memory preserved Retrograde amnesia (RA) –Ribot gradients Pattern of correlations between AA and RA –No perfect correlation between AA and RA

x retrograde amnesia anterograde amnesia lesionpresentpast Normal forgetting

An example of retrograde amnesia patient data Kopelman (1989) News events test

Retrograde amnesia Primary cause: loss of links Ribot gradients Shrinkage

Anterograde amnesia Primary cause: loss of modulatory system Secondary cause: loss of links Preserved implicit memory

Semantic dementia The term was adopted recently to describe a new form of dementia, notably by Julie Snowden et al. (1989, 1994) and by John Hodges et al. (1992, 1994) Semantic dementia is almost a mirror- image of amnesia

Neuropsychology of semantic dementia Progressive loss of semantic knowledge Word-finding problems Comprehension difficulties No problems with new learning Lesions mainly located in the infero-lateral temporal cortex but (early in the disease) with sparing of the hippocampus

Severe loss of trace connections Stage-2 learning proceeds as normal Stage 3 learning strongly impaired Non-rehearsed memories will be lost No consolidation in semantic dementia

Semantic dementia in TraceLink Primary cause: loss of trace-trace connections Stage-3 (and 4) memories cannot be formed: no consolidation The preservation of new memories will be dependent on constant rehearsal

Connectionist implementation of the TraceLink model With Martijn Meeter from the University of Amsterdam

Some details of the model 42 link nodes, 200 trace nodes for each pattern –7 nodes are active in the link system –10 nodes in the trace system Trace system has lower learning rate that the link system

How the simulations work: One simulated ‘day’ A new pattern is activated The pattern is learned Because of low learning rate, the pattern is not well encoded at first in the trace system A period of ‘simulated dreaming’ follows –Nodes are activated randomly by the model –This random activity causes recall of a pattern –A recalled pattern is than learned extra

(Patient data) Kopelman (1989) News events test

A simulation with TraceLink

Frequency of consolidation of patterns over time

Strongly and weakly encoded patterns Mixture of weak, middle and strong patterns Strong patterns had a higher learning parameter (cf. longer learning time)

Transient Global Amnesia (TGA) (Witnessed onset) of severe anterograde and retrograde amnesia Resolves within 24 hours Retrograde amnesia may have Ribot gradients Hippocampal area is most probably implicated

Transient Global Amnesia (TGA)

Other simulations Focal retrograde amnesia Levels of processing Semantic dementia Implicit memory More subtle lesions (e.g., only within-link connections, cf. CA1 lesions)

The Memory Chain Model: a very abstract neural network With Antonio Chessa from the University of Amsterdam

Abstracting TraceLink (level 1) Model formulated within the mathematical framework of point processes Generalizes TraceLink’s two-store approach to multiple neural ‘stores’ –trace system –link system –working memory, short-term memory, etc. A store corresponds to a neural process or structure

Learning and forgetting as a stochastic process: 1-store example A recall cue (e.g., a face) may access different aspects of a stored memory If a point is found in the neural cue area, the correct response (e.g., the name) can be given Learning ForgettingSuccessful Recall Unsuccessful Recall

Neural network interpretation Jo Brand

Single-store point process The expected number of points in the cue area after learning is called  This  is directly increased by learning and also by more effective cueing At each time step, points die The probability of survival of a point is denoted by a Link systemRetrieval  a Survival probability

Some aspects of the point process model Model of simultaneous learning and forgetting Clear relationship between signal detection theory (d'), recall (p), savings (Ebbinghaus’ Q), and Crovitz-type distribution functions Multi-trial learning and multi-trial savings Currently applied to over 250 experiments in learning and forgetting, since 1885

Forgetting curve If we need to find at least one point we obtain the following curve (one-store case): We predict a flex point when the initial recall is at least  is the intensity of the process (expected number of points) and a is the decay parameter

Example: Single-store model fitted to short-term forgetting data R 2 = 0,985

Flex points versus initial retention level: an analysis of 200 data sets 0.63: ‘overlearning threshold’

Multi-store generalization Information about the current event passes through many neural ‘stores’ The retina, for example, holds a lot of information very briefly The cerebral cortex holds very little information (of the current event) for a very long time

General principles of the PPM multi-store model A small part of the information is passed to the next store before it decays completely Subsequent stores hold information for longer time periods: slower decay rates in ‘higher’ stores

Two-store model While neural store 1 is decaying (with rate a 1 ) it induces new points (representations) in store 2 Induction rate is linear with the intensity in store 1 and has induction rate  2 The points in store immediately start to decay as well (at a lower rate a 2 )

Example of two neural stores Store 1: firing neural groups Store 2: synaptic connections between the neural groups Other interpretation are possible as well, e.g.: –Store 1: hippocampus –Store 2: cerebral cortex Skip

Example of two neural stores: encoding phase H Store 1 Additional cue area Stimulus

Storage phase: decay of neural groups and Hebbian learning Store 2

?R Q Recall phase: retrieval through cue Q Skip

Decomposition of intensity  (t) into encoding, storage, and retrieval

The contributions of S individual neural stores can simply be added

Two-store model retention function: r 12 (t)= r 1 (t)+ r 2 (t)

The retention function for the third-store of a three-store model

Recall probability p(t) as a function of different learning times l is the learning rate l is the learning time r(t) is the decline function t time since learning

Saturation assumption

Hellyer (1962). Recall as a function of 1, 2, 4 and 8 presentations Two-store model with saturation. Parameters are  1 = 7.4, a 1 = 0.53,  2 = 0.26, a 2 = 0.31, r max = 85; R 2 =.986 Skip

Amnesia: animal data Retrograde amnesia

Cho & Kesner (1996). (mice) R 2 =0.96

Summary of animal data

Frankland et al. (2001) study  -CaMKB-dependent plasticity (in neocortex) switched off in knock-out mice No LTP measurable in neocortex but LTP in hippocampus was largely normal Forgetting curves with different levels of initial learning were measured A learning curve was measured Assumption: use r 1[2] (t) for knock-out mice

Forgetting after 3 shocks, using three parameters

Using the same three parameters and a massed-learning correction.

Controls receive 1 shock, experimental animals 3 shocks (no new free parameters).

Repeated learning for experimental animals (no new free parameters)

Summary of ‘cortical amnesia’. Using only 4 parameters for all curves (R 2 = 0.976).

Amnesia: human data

Application to retrograde amnesia Data on clinical tests cannot be used for direct modeling The reason is that remote time periods in these tests are typically made easier Data for the different time periods are therefore not equivalent Our model may offer a solution here: the relative retrograde gradient or rr-gradient

Sometimes this problems occurs with animal data as well Wiig, Cooper, and Bear (1996) Used non-counterbalanced stimuli

Wiig, Cooper & Bear (1996). (rats) R 2 =0.28

Wiig, Cooper & Bear (1996). (rats) with rr-gradient: R 2 =0.84

Define the relative retrograde gradient or rr-gradient

rr-gradient (continued)

The rr-gradient does not have parameters for learning strength  1 or cue strength q

Recall probability p(t) must transformed to retention r(t)

Albert et al. (1979), naming of famous faces

Squire, Haist, and Shimamura (1989), recall of public events

Concluding remarks In this presentation, we have shown models at two levels of abstraction: –Mathematical, based on point processes –Computational, based on simplified neural networks

Concluding remarks These models incorporate data from: –Neuroanatomy and neurophysiology –Neurology and neuropsychology –Experimental psychology The aim is to integrate these various sources of data into a single theory that is implemented in a series of coordinated models

Concluding remarks Given that the brain is exceedingly complex, we need models at various levels of abstraction to aid our understanding This is especially true when trying to unravel the link between the brain and human behavior, which is extremely complex itself Hence, models are of particular use in the new, interdisciplinary field of cognitive neuroscience