episodic memory: why and how?

95
MÁTÉ LENGYEL Computational and Biological Learning Lab Department of Engineering University of Cambridge EPISODIC MEMORY: WHY AND HOW? (THE POWERS AND PERILS OF BAYESIAN INFERENCE IN THE BRAIN)

Upload: others

Post on 20-Dec-2021

10 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: EPISODIC MEMORY: WHY AND HOW?

MÁTÉ LENGYEL

Computational and Biological Learning LabDepartment of Engineering

University of Cambridge

EPISODIC MEMORY: WHY AND HOW?(THE POWERS AND PERILS OF BAYESIAN INFERENCE IN THE BRAIN)

Page 2: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

EPISODIC MEMORY: AN EXAMPLE

2

I raised to my lips a spoonful of the tea in which I had soaked a morsel of the cake. ... And suddenly the memory returns. The taste was that of the little crumb of madeleine which on Sunday mornings at Combray, when I went to say good day to her in her bedroom, my aunt Léonie used to give me, dipping it first in her own cup of real or of lime-flower tea.

Marcel Proust: À la recherche du temps perdu

Page 3: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

EPISODIC MEMORY: AN EXAMPLE

2

I raised to my lips a spoonful of the tea in which I had soaked a morsel of the cake. ... And suddenly the memory returns. The taste was that of the little crumb of madeleine which on Sunday mornings at Combray, when I went to say good day to her in her bedroom, my aunt Léonie used to give me, dipping it first in her own cup of real or of lime-flower tea.

Marcel Proust: À la recherche du temps perdu

PART I: WHY DO WE HAVE SUCH MEMORIES?! specific personal experiences! organised into sequences of events

Page 4: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

EPISODIC MEMORY: AN EXAMPLE

2

I raised to my lips a spoonful of the tea in which I had soaked a morsel of the cake. ... And suddenly the memory returns. The taste was that of the little crumb of madeleine which on Sunday mornings at Combray, when I went to say good day to her in her bedroom, my aunt Léonie used to give me, dipping it first in her own cup of real or of lime-flower tea.

Marcel Proust: À la recherche du temps perdu

PART I: WHY DO WE HAVE SUCH MEMORIES?! specific personal experiences! organised into sequences of events

PART II: HOW DOES THIS HAPPEN?! memories laid down in the past ! recalled in response to a cue

Page 5: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

ON THE USE OF MEMORIES

3

Page 6: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

ON THE USE OF MEMORIES

3

ah, those nice days back in Combray

Page 7: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

ON THE USE OF MEMORIES

3

experiencedata

predictions

ah, those nice days back in Combray

representation in memorysufficient statistics

Page 8: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

ON THE USE OF MEMORIES

3

experiencedata

predictions

if I soak my cake in my tea ! it will taste good

representation in memorysufficient statistics

Page 9: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

ON THE USE OF MEMORIES

3

experiencedata

predictions

if I soak my cake in my tea ! it will taste good

model of the environmentsufficient statistics

‘semantic’ memory:

Page 10: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

ON THE USE OF MEMORIES

3

experiencedata

predictions

‘episodic’ memory:select episodes

data points

if I soak my cake in my tea ! it will taste good

model of the environmentsufficient statistics

‘semantic’ memory:

Page 11: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

ON THE USE OF MEMORIES

3

experiencedata

predictions

‘episodic’ memory:select episodes

data points?if I soak my cake in my tea !

it will taste good

model of the environmentsufficient statistics

‘semantic’ memory:

Page 12: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

ON THE USE OF MEMORIES

3

experiencedata

‘episodic’ memory:select episodes

data points

planning for the futuresequential decision making

?if I soak my cake in my tea !

it will taste good

model of the environmentsufficient statistics

‘semantic’ memory:

Page 13: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

ON THE USE OF MEMORIES

3

experiencedata

‘episodic’ memory:select episodes

data points

planning for the futuresequential decision making

?

what shall I do next to taste something nice in the end?

model of the environmentsufficient statistics

‘semantic’ memory:

Page 14: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

ON THE USE OF MEMORIES

3

experiencedata

‘episodic’ memory:select episodes

data points

planning for the futuresequential decision making

?

what shall I do next to taste something nice in the end?

model of the environmentsufficient statistics

‘semantic’ memory:

! delayed rewards! temporal credit

assignment

Page 15: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

ON THE USE OF MEMORIES

3

experiencedata

learning control

‘episodic’ memory:select episodes

data points

planning for the futuresequential decision making

?

what shall I do next to taste something nice in the end?

model of the environmentsufficient statistics

‘semantic’ memory:

! delayed rewards! temporal credit

assignment

Page 16: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

ON THE USE OF MEMORIES

3

experiencedata

learning control

‘episodic’ memory:select episodes

data points

planning for the futuresequential decision making

?

what shall I do next to taste something nice in the end?

model of the environmentsufficient statistics

‘semantic’ memory:

clever hard

! delayed rewards! temporal credit

assignment

Page 17: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

ON THE USE OF MEMORIES

3

experiencedata

learning control

‘episodic’ memory:select episodes

data points

planning for the futuresequential decision making

?

what shall I do next to taste something nice in the end?

model of the environmentsufficient statistics

‘semantic’ memory:

clever

dull

hard

easy! delayed rewards! temporal credit

assignment

Page 18: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel 4

SEQUENTIAL DECISION-MAKING UNDER UNCERTAINTYLengyel & Dayan, NIPS 2007

Page 19: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel 4

SEQUENTIAL DECISION-MAKING UNDER UNCERTAINTYLengyel & Dayan, NIPS 2007

Page 20: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

SEMANTIC MEMORY: MODEL-BASED CONTROL

5

Lengyel & Dayan, NIPS 2007

Page 21: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

SEMANTIC MEMORY: MODEL-BASED CONTROL

5

! learns a model of the environment(posterior distribution over parameters, etc)

Lengyel & Dayan, NIPS 2007

Page 22: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

SEMANTIC MEMORY: MODEL-BASED CONTROL

5

T=0

! learns a model of the environment(posterior distribution over parameters, etc)

Lengyel & Dayan, NIPS 2007

Page 23: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

SEMANTIC MEMORY: MODEL-BASED CONTROL

5

T=0T=1

! learns a model of the environment(posterior distribution over parameters, etc)

Lengyel & Dayan, NIPS 2007

Page 24: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

SEMANTIC MEMORY: MODEL-BASED CONTROL

5

T=0T=1T=100

! learns a model of the environment(posterior distribution over parameters, etc)

Lengyel & Dayan, NIPS 2007

Page 25: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

SEMANTIC MEMORY: MODEL-BASED CONTROL

5

T=0T=1T=100

! learns a model of the environment(posterior distribution over parameters, etc)

! selects actions by recursive ‘mental simulation’

Lengyel & Dayan, NIPS 2007

Page 26: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

SEMANTIC MEMORY: MODEL-BASED CONTROL

5

T=0T=1T=100

! learns a model of the environment(posterior distribution over parameters, etc)

! selects actions by recursive ‘mental simulation’

! tree search implies combinatorial explosion

Lengyel & Dayan, NIPS 2007

Page 27: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

SEMANTIC MEMORY: MODEL-BASED CONTROL

5

T=0T=1T=100

! learns a model of the environment(posterior distribution over parameters, etc)

! selects actions by recursive ‘mental simulation’

! tree search implies combinatorial explosion

! approximations are necessary

Lengyel & Dayan, NIPS 2007

Page 28: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

SEMANTIC MEMORY: MODEL-BASED CONTROL

5

T=0T=1T=100

! learns a model of the environment(posterior distribution over parameters, etc)

! selects actions by recursive ‘mental simulation’

! tree search implies combinatorial explosion

! approximations are necessary

! effective computational noise

Lengyel & Dayan, NIPS 2007

Page 29: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

EPISODIC MEMORY: MODEL-FREE CONTROL

6

Lengyel & Dayan, NIPS 2007

Page 30: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

EPISODIC MEMORY: MODEL-FREE CONTROL

6

! stores specific episodes retrospectively(state—action—…—reward sequences)

Lengyel & Dayan, NIPS 2007

Page 31: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

EPISODIC MEMORY: MODEL-FREE CONTROL

6

T=0! stores specific episodes retrospectively

(state—action—…—reward sequences)

Lengyel & Dayan, NIPS 2007

Page 32: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

EPISODIC MEMORY: MODEL-FREE CONTROL

6

T=0T=1! stores specific episodes retrospectively

(state—action—…—reward sequences)

Lengyel & Dayan, NIPS 2007

Page 33: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

EPISODIC MEMORY: MODEL-FREE CONTROL

6

T=0T=1T=100! stores specific episodes retrospectively

(state—action—…—reward sequences)

Lengyel & Dayan, NIPS 2007

Page 34: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

EPISODIC MEMORY: MODEL-FREE CONTROL

6

T=0T=1T=100! stores specific episodes retrospectively

(state—action—…—reward sequences)

! selects action that yielded maximalultimate reward in past episodes startingfrom current state

Lengyel & Dayan, NIPS 2007

Page 35: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

EPISODIC MEMORY: MODEL-FREE CONTROL

6

T=0T=1T=100! stores specific episodes retrospectively

(state—action—…—reward sequences)

! selects action that yielded maximalultimate reward in past episodes startingfrom current state

compatible with:

! hippocampal involvement in" processing sequential memories

(Fortin et al, 2002; Ergorul & Eichenbaum, 2006; Manns et al, 2007, Lehn et al, 2009)

" imagining new experiences(Hassabis et al, 2007)

! awake forward replay at decision points (Johnson & Redish, 2007)

! reward and episodic information integrated (Lansink et al, 2009; Rossato et al, 2009)

Lengyel & Dayan, NIPS 2007

Page 36: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

COMPARING THE TWO SYSTEMS

7

Lengyel & Dayan, NIPS 2007

Page 37: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

COMPARING THE TWO SYSTEMS

7

• vary complexity of the environment (A, B, and D)

Lengyel & Dayan, NIPS 2007

Page 38: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

COMPARING THE TWO SYSTEMS

7

• vary complexity of the environment (A, B, and D)

number of actionsA

Lengyel & Dayan, NIPS 2007

Page 39: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

COMPARING THE TWO SYSTEMS

7

• vary complexity of the environment (A, B, and D)

number of actionsA branching factor

B

Lengyel & Dayan, NIPS 2007

Page 40: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

COMPARING THE TWO SYSTEMS

7

• vary complexity of the environment (A, B, and D)

depthD

number of actionsA branching factor

B

Lengyel & Dayan, NIPS 2007

Page 41: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

COMPARING THE TWO SYSTEMS

7

• vary complexity of the environment (A, B, and D)

• vary amount of experience available to semantic and episodic memory systems

depthD

number of actionsA branching factor

B

Lengyel & Dayan, NIPS 2007

Page 42: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

COMPARING THE TWO SYSTEMS

7

• vary complexity of the environment (A, B, and D)

• vary amount of experience available to semantic and episodic memory systems

• compute average performance of three systems

depthD

number of actionsA branching factor

B

Lengyel & Dayan, NIPS 2007

Page 43: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

COMPARING THE TWO SYSTEMS

7

• vary complexity of the environment (A, B, and D)

• vary amount of experience available to semantic and episodic memory systems

• compute average performance of three systems

1. perfect semantic memory-based control ! theoretical upper bound

depthD

number of actionsA branching factor

B

Lengyel & Dayan, NIPS 2007

Page 44: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

COMPARING THE TWO SYSTEMS

7

• vary complexity of the environment (A, B, and D)

• vary amount of experience available to semantic and episodic memory systems

• compute average performance of three systems

1. perfect semantic memory-based control ! theoretical upper bound2. approximate semantic memory-based controller

depthD

number of actionsA branching factor

B

Lengyel & Dayan, NIPS 2007

Page 45: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

COMPARING THE TWO SYSTEMS

7

• vary complexity of the environment (A, B, and D)

• vary amount of experience available to semantic and episodic memory systems

• compute average performance of three systems

1. perfect semantic memory-based control ! theoretical upper bound2. approximate semantic memory-based controller 3. episodic memory-based controller

depthD

number of actionsA branching factor

B

Lengyel & Dayan, NIPS 2007

Page 46: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

THE PERFECT MODEL-BASED SYSTEM

8

V ′1

V ′2

V ′B

!

V

!

Q1

QA

p11

pAB

core idea: analysing a sub-treelet

Lengyel & Dayan, NIPS 2007

Page 47: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

THE PERFECT MODEL-BASED SYSTEM

8

V ′1

V ′2

V ′B

!

V

!

Q1

QA

p11

pAB

core idea: analysing a sub-treelet

Lengyel & Dayan, NIPS 2007

value of state

Page 48: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

THE PERFECT MODEL-BASED SYSTEM

8

V ′1

V ′2

V ′B

!

V

!

Q1

QA

p11

pAB

core idea: analysing a sub-treelet

Lengyel & Dayan, NIPS 2007

values of available action

Page 49: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

THE PERFECT MODEL-BASED SYSTEM

8

V ′1

V ′2

V ′B

!

V

!

Q1

QA

p11

pAB

core idea: analysing a sub-treelet

Lengyel & Dayan, NIPS 2007

values of successor states

Page 50: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

THE PERFECT MODEL-BASED SYSTEM

8

V ′1

V ′2

V ′B

!

V

!

Q1

QA

p11

pAB

core idea: analysing a sub-treelet

Lengyel & Dayan, NIPS 2007

µV ′

σV ′

Page 51: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

THE PERFECT MODEL-BASED SYSTEM

8

V ′1

V ′2

V ′B

!

V

!

Q1

QA

p11

pAB

core idea: analysing a sub-treelet

Lengyel & Dayan, NIPS 2007

µV ′

σV ′

averaging

Page 52: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

THE PERFECT MODEL-BASED SYSTEM

8

V ′1

V ′2

V ′B

!

V

!

Q1

QA

p11

pAB

core idea: analysing a sub-treelet

Lengyel & Dayan, NIPS 2007

µV ′

σV ′

µQ

σQ

averaging

CQ

Page 53: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

THE PERFECT MODEL-BASED SYSTEM

8

V ′1

V ′2

V ′B

!

V

!

Q1

QA

p11

pAB

core idea: analysing a sub-treelet

Lengyel & Dayan, NIPS 2007

µV ′

σV ′

µQ

σQ

averagingmax

CQ

Page 54: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

THE PERFECT MODEL-BASED SYSTEM

8

V ′1

V ′2

V ′B

!

V

!

Q1

QA

p11

pAB

core idea: analysing a sub-treelet

Lengyel & Dayan, NIPS 2007

µV ′

σV ′

µQ

σQ

µV

σV

averagingmax

CQ

Page 55: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

THE PERFECT MODEL-BASED SYSTEM

8

V ′1

V ′2

V ′B

!

V

!

Q1

QA

p11

pAB

core idea: analysing a sub-treelet whole environment

Lengyel & Dayan, NIPS 2007

µV ′

σV ′

µQ

σQ

µV

σV

averagingmax

CQ

Page 56: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

THE PERFECT MODEL-BASED SYSTEM

8

more complex environment ! more potential for high rewards

V ′1

V ′2

V ′B

!

V

!

Q1

QA

p11

pAB

core idea: analysing a sub-treelet whole environment

Lengyel & Dayan, NIPS 2007

µV ′

σV ′

µQ

σQ

µV

σV

averagingmax

CQ

Page 57: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

THE EFFECTS OF APPROXIMATIONS

9

true action values ! noisy versions

Q̃ = η1Q + η2z z ∼ N (0, 1)

Lengyel & Dayan, NIPS 2007

Page 58: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

THE EFFECTS OF APPROXIMATIONS

9

true action values ! noisy versions

Q̃ = η1Q + η2z z ∼ N (0, 1)

Lengyel & Dayan, NIPS 2007

Q1 Q2

Page 59: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

THE EFFECTS OF APPROXIMATIONS

9

true action values ! noisy versions

Q̃ = η1Q + η2z z ∼ N (0, 1)

Lengyel & Dayan, NIPS 2007

P(Q̃2 | Q2

)P

(Q̃1 | Q1

)

Q1 Q2

Page 60: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

THE EFFECTS OF APPROXIMATIONS

9

true action values ! noisy versions

Q̃ = η1Q + η2z z ∼ N (0, 1)actual reward dependson Q of the actionwith highest Q̃

Lengyel & Dayan, NIPS 2007

P(Q̃2 | Q2

)P

(Q̃1 | Q1

)

Q1 Q2

Page 61: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

THE EFFECTS OF APPROXIMATIONS

9

true action values ! noisy versions

Q̃ = η1Q + η2z z ∼ N (0, 1)

noise-to-signal ratio: ω2 =η22

η21 σ2

Q

actual reward dependson Q of the actionwith highest Q̃

Lengyel & Dayan, NIPS 2007

P(Q̃2 | Q2

)P

(Q̃1 | Q1

)

Q1 Q2

Page 62: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

THE EFFECTS OF APPROXIMATIONS

9

true action values ! noisy versions

Q̃ = η1Q + η2z z ∼ N (0, 1)

noise-to-signal ratio: ω2 =η22

η21 σ2

Q

actual reward dependson Q of the actionwith highest Q̃

Lengyel & Dayan, NIPS 2007

σQ

P(Q̃2 | Q2

)P

(Q̃1 | Q1

)

Q1 Q2

ω ω

Page 63: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

THE EFFECTS OF APPROXIMATIONS

9

single state

true action values ! noisy versions

Q̃ = η1Q + η2z z ∼ N (0, 1)

noise-to-signal ratio: ω2 =η22

η21 σ2

Q

actual reward dependson Q of the actionwith highest Q̃

Lengyel & Dayan, NIPS 2007

σQ

P(Q̃2 | Q2

)P

(Q̃1 | Q1

)

Q1 Q2

ω ω

Page 64: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

THE EFFECTS OF APPROXIMATIONS

9

incre

asin

g le

vels o

fco

mputa

tional n

oise

single state whole environment

true action values ! noisy versions

Q̃ = η1Q + η2z z ∼ N (0, 1)

noise-to-signal ratio: ω2 =η22

η21 σ2

Q

actual reward dependson Q of the actionwith highest Q̃

Lengyel & Dayan, NIPS 2007

σQ

P(Q̃2 | Q2

)P

(Q̃1 | Q1

)

Q1 Q2

ω ω

Page 65: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

THE EFFECTS OF APPROXIMATIONS

9

incre

asin

g le

vels o

fco

mputa

tional n

oise

deeper environment ! noise is more deleterious

single state whole environment

true action values ! noisy versions

Q̃ = η1Q + η2z z ∼ N (0, 1)

noise-to-signal ratio: ω2 =η22

η21 σ2

Q

actual reward dependson Q of the actionwith highest Q̃

Lengyel & Dayan, NIPS 2007

σQ

P(Q̃2 | Q2

)P

(Q̃1 | Q1

)

Q1 Q2

ω ω

Page 66: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

! "!! #!! $!! %!! &!!!

!'&

"

!"

! "!! #!! $!! %!! &!!!

!'"

!'#

!#

! "!! #!! $!! %!! &!!!

"!

#!

$!

()*+,-!./!+/)0123,.)/

4!

THE EFFECT OF LEARNING

10

key idea: ignorance about the environment ! additional noise

Lengyel & Dayan, NIPS 2007

Page 67: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

! "!! #!! $!! %!! &!!!

!'&

"

!"

! "!! #!! $!! %!! &!!!

!'"

!'#

!#

! "!! #!! $!! %!! &!!!

"!

#!

$!

()*+,-!./!+/)0123,.)/

4!

THE EFFECT OF LEARNING

10

key idea: ignorance about the environment ! additional noise

number of times each state-action pair

is visited

Lengyel & Dayan, NIPS 2007

Page 68: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

! "!! #!! $!! %!! &!!!

!'&

"

!"

! "!! #!! $!! %!! &!!!

!'"

!'#

!#

! "!! #!! $!! %!! &!!!

"!

#!

$!

()*+,-!./!+/)0123,.)/

4!

THE EFFECT OF LEARNING

10

key idea: ignorance about the environment ! additional noise

number of times each state-action pair

is visited

time requiredto collect it

∝ A BD

Lengyel & Dayan, NIPS 2007

Page 69: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

THE EFFECTS OF COMPUTATIONAL + IGNORANCE NOISE

11

Lengyel & Dayan, NIPS 2007

Page 70: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

THE EFFECTS OF COMPUTATIONAL + IGNORANCE NOISE

11

incre

asin

g le

vels o

fco

mputa

tional n

oise

Lengyel & Dayan, NIPS 2007

Page 71: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

THE EFFECTS OF COMPUTATIONAL + IGNORANCE NOISE

11

approximations: more adverse effects early in learning!

incre

asin

g le

vels o

fco

mputa

tional n

oise

Lengyel & Dayan, NIPS 2007

Page 72: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

THE EFFECTS OF COMPUTATIONAL + IGNORANCE NOISE

11

approximations: more adverse effects early in learning!

incre

asin

g le

vels o

fco

mputa

tional n

oise

Lengyel & Dayan, NIPS 2007

Page 73: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

THE EFFECTS OF COMPUTATIONAL + IGNORANCE NOISE

11

approximations: more adverse effects early in learning!

incre

asin

g le

vels o

fco

mputa

tional n

oise

room for alternative decision making systems in low-data limit

Lengyel & Dayan, NIPS 2007

Page 74: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

EPISODIC VS. SEMANTIC MEMORY-BASED DECISION MAKING

12

Lengyel & Dayan, NIPS 2007

Page 75: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

EPISODIC VS. SEMANTIC MEMORY-BASED DECISION MAKING

12

episodic advantage early in learning

Lengyel & Dayan, NIPS 2007

Page 76: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

EPISODIC VS. SEMANTIC MEMORY-BASED DECISION MAKING

12

incre

asin

g e

nviro

nm

enta

l com

ple

xity

episodic advantage early in learning

Amount of experience

Lengyel & Dayan, NIPS 2007

Page 77: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

EPISODIC VS. SEMANTIC MEMORY-BASED DECISION MAKING

12

incre

asin

g e

nviro

nm

enta

l com

ple

xity

episodic advantage early in learning

lasts longer for more complex environments

Amount of experience

Lengyel & Dayan, NIPS 2007

Page 78: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

SUMMARY

13

Page 79: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

envi

ronm

enta

l co

mple

xit

y

learning

SUMMARY

13

Page 80: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

envi

ronm

enta

l co

mple

xit

y

learning

SUMMARY

13

semantic memory

Page 81: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

envi

ronm

enta

l co

mple

xit

y

learning

SUMMARY

13

semantic memoryneocortex

Page 82: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

envi

ronm

enta

l co

mple

xit

y

learning

SUMMARY

13

semantic memory

episodic memory

neocortex

hippocampus

Page 83: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

envi

ronm

enta

l co

mple

xit

y

learning

SUMMARY

13

semantic memory

episodic memory

‘value’ or procedural memory

neocortex

hippocampus

striatum

Page 84: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

envi

ronm

enta

l co

mple

xit

y

learning

SUMMARY

13

semantic memory

episodic memory

‘value’ or procedural memory

‘consolidation’

neocortex

hippocampus

striatum

Page 85: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

envi

ronm

enta

l co

mple

xit

y

learning

SUMMARY

13

semantic memory

episodic memory

‘value’ or procedural memory

‘consolidation’

consolidation: transfer of control rather than transfer of memories?

neocortex

hippocampus

striatum

Page 86: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

envi

ronm

enta

l co

mple

xit

y

learning

SUMMARY

13

semantic memory

episodic memory

‘value’ or procedural memory

‘consolidation’

‘competing memory systems’

consolidation: transfer of control rather than transfer of memories?

neocortex

hippocampus

striatum

Page 87: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

envi

ronm

enta

l co

mple

xit

y

learning

SUMMARY

13

semantic memory

episodic memory

‘value’ or procedural memory

‘consolidation’

‘habitization’

‘competing memory systems’

Daw et al, 2005

consolidation: transfer of control rather than transfer of memories?

neocortex

hippocampus

striatum

Page 88: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

envi

ronm

enta

l co

mple

xit

y

learning

SUMMARY

13

semantic memory

episodic memory

‘value’ or procedural memory

‘consolidation’

‘habitization’

‘competing memory systems’

Daw et al, 2005

consolidation: transfer of control rather than transfer of memories?

neocortex

hippocampus

striatum

environmental non-stationarity

Page 89: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

OPEN QUESTIONS

14

Page 90: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

OPEN QUESTIONS

14

! granularity of integration of episodic with model-based system

Page 91: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

OPEN QUESTIONS

14

! granularity of integration of episodic with model-based system

! episodes store outcomes (goal-directed) or rewards (habitual)

Page 92: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

OPEN QUESTIONS

14

! granularity of integration of episodic with model-based system

! episodes store outcomes (goal-directed) or rewards (habitual)

! arbitration between parallel systems

Page 93: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

OPEN QUESTIONS

14

! granularity of integration of episodic with model-based system

! episodes store outcomes (goal-directed) or rewards (habitual)

! arbitration between parallel systems! need to represent uncertainty (Daw et al, 2005)

Page 94: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

OPEN QUESTIONS

14

! granularity of integration of episodic with model-based system

! episodes store outcomes (goal-directed) or rewards (habitual)

! arbitration between parallel systems! need to represent uncertainty (Daw et al, 2005)

! replay during sleep not for consolidation

Page 95: EPISODIC MEMORY: WHY AND HOW?

Máté Lengyel: Episodic memory: why and how? BCCN 2009, 3 October 2009 http://www.eng.cam.ac.uk/~m.lengyel

OPEN QUESTIONS

14

! granularity of integration of episodic with model-based system

! episodes store outcomes (goal-directed) or rewards (habitual)

! arbitration between parallel systems! need to represent uncertainty (Daw et al, 2005)

! replay during sleep not for consolidation ! keeping memory representations in register (Káli & Dayan, 2005)