wrightteam

Information about wrightteam

Published on February 7, 2008

Author: Veronica1

Source: authorstream.com

Content

Synchrony and Learning in an Olfaction-Inspired Network Model: Synchrony and Learning in an Olfaction-Inspired Network Model Group Members: Wally Ian Mike Corrie Jeff Cindy Group Advisors: Geri Wright David Terman Introduction to Olfaction: Introduction to Olfaction The main task of the olfactory system is to detect thousands of odor molecules Odors are multidimensional, but we generally perceive them as a unit. General Question: how does the olfactory system solve such a complicated pattern-learning problem Functional Anatomy: Functional Anatomy From Firestein, Nature 2001 Spatial Code: Spatial Code Identical sharply tuned receptor neurons track an odor through “labelled lines” Sharply tuned receptors Glomeruli of the antennal lobe Adapted from Firestein, 2001 and http://www.neurobiologie.fu-berlin.de/galizia Temporal Code: Temporal Code From Laurent 2002 Outline: Description of Our Olfactory-Like Computational Model Description of Our Synaptic Learning Mechanism Popcorn and Pearl Effect and the Relation Between Synchrony and Gampa and Gadap. Spatial Representations of Odor Stimuli Learning Behaviors of Our Model Possibilities for Future Direction Outline Basic Network Architecture: Basic Network Architecture Olfactory Sensory Neuron Mitral Cell Granular Cell Slide 8: Architecture of Our Model: MatLab One-Dimensional Array of ‘N’ Identical Neurons Modeled by Reduced Hodgkin-Huxley Type Equations Gaussian Weighted Synaptic Connections Global or Selective Applied Current Cell Number Time Step Time Step Avg. Membrane Voltage Slide 9: Excitatory and Inhibitory Footprints in Our Model Play a Crucial Role in Synaptic Learning and Spatial Resolution: Our model has local excitation and distant inhibition represented mathematically by Gaussian functions. x Slide 10: Metabotropic Process: A model of this type can be thought of biologically as a form of synaptic facilitation mediated by a metabotropic receptor and carried out by a second messenger cascade. Slide 11: Mechanism of Learning, Adapting Current: Introduction of a new synaptic gating variable which grows with pre-synaptic firing and decays much more slowly then the other synaptic variable. The “Popcorn” and “Pearl Effect”: The “Popcorn” and “Pearl Effect” When no synchronous excitatory nerve activity exists in a network of cells in the model, the movement between cells is a chaotic effect among cells, also known as the “Popcorn Effect”. Nerves are firing at different rates to a stimuli, such as an odor in the Olfactory system. When synchronous excitatory nerve activity exists in a network of cells in the model, the movement between cells creates a wave or an exact jump effect among cells, also known as the “Pearl Effect”. All nerves are firing at the same rate to a stimuli, such as an odor in the Olfactory system. Show Video of Popcorn and Pearl effect from this slide Chalkboard Comparison of Graphs: Chalkboard Comparison of Graphs Line graphs have time on the x axis and average cell voltage on the y-axis Graphs along the horizontal ( from left to right) compare a change in g-adap with variables set from .1 (the left side) to 10 (the right side) Graphs along the vertical (from bottom to top) compare a change in gampa with variables set at .01 (the bottom) to .08 (the top) There is no change in gadap if gampa is zero, causing the “Popcorn Effect” There is a change in gadap if gampa is increased, causing the “Pearl Effect” Gampa at .01 and gadap at 0 The “Popcorn Effect” with unsynchronized movement: Gampa at .01 and gadap at 0The “Popcorn Effect” with unsynchronized movement Gampa at .08 and gadap at 10 The “Pearl Effect” with synchronized movement: Gampa at .08 and gadap at 10The “Pearl Effect” with synchronized movement Example of a 50:50 Popcorn-Pearl Effect”: Example of a 50:50 Popcorn-Pearl Effect” Determining the “Popcorn and Pearl” threshold curve: Determining the “Popcorn and Pearl” threshold curve Using a 50:50 popcorn to Pearl pattern, an X,Y line of the gampa-gadap relationship was graphed The pattern appeared to follow either an exponential or power regression The power regression was selected as a result of a higher correlation coefficient and a more realistic picture of the model (no zero value on the Y) Using the 50:50 curve, predictions were made and verified with other points Prediction 1 - Values of gampa at .015 and gadap at 2.6 showing 50:50 “Popcorn-Pearl Effect”: Prediction 1 - Values of gampa at .015 and gadap at 2.6showing 50:50 “Popcorn-Pearl Effect” Prediction 2 - Values of Gampa at .01 and gadap at 6.6 showing 50:50 “Popcorn-Pearl Effect”: Prediction 2 - Values of Gampa at .01 and gadap at 6.6showing 50:50 “Popcorn-Pearl Effect” Determining the “Popcorn and Pearl” threshold curve: Determining the “Popcorn and Pearl” threshold curve Using a 50:50 popcorn to Pearl pattern, an X,Y line of the gampa-gadap relationship was graphed The pattern appeared to follow either an exponential or power regression The power regression was selected as a result of a higher correlation coefficient and a more realistic picture of the model (no zero value on the Y) Using the 50:50 curve, predictions were made and verified with other points A second curve was formed using threshold values of Pearling which also followed a power curve. The area below the second curve represents settings where Pearling should be improbable Points where selected for verification Below Threshold Values of Gampa at .01 and gadap at 1.78 showing “popcorn Effect”: Below Threshold Values of Gampa at .01 and gadap at 1.78showing “popcorn Effect” Below Threshold Values of Gampa at .021 and gadap at .33 showing “Popcorn Effect”: Below Threshold Values of Gampa at .021 and gadap at .33showing “Popcorn Effect” Spatial Coding of Odor Signals : Spatial Coding of Odor Signals From Sachse and Galizia, 2001 Bee Antennal Lobe Image Motivation: If odors are spatially represented, what influences the ability to discriminate two distinct odors? The Model: The Model Similar model to previous, but with localized stimuli The question ?: The question ? For three different stimulus locations (arrows), we considered the effects of changes to : G-ampa: Excitatory conductance of synapses G-gaba: Inhibitory conductance of synapses How does the network respond when presented with two external stimuli? Observed Trends As Excitatory Synaptic Conductance Increases:: Observed Trends As Excitatory Synaptic Conductance Increases: Network displays amplification: Higher signal strength Greater recruitment over network Time Cell number voltage Observed Trends As Inhibitory Synaptic Conductance Increases:: Observed TrendsAs Inhibitory Synaptic Conductance Increases: Network Displays Resolution: Greater signal resolution PIR induced oscillation Time Voltage Cell Number Resolution of Distinct Signals: Resolution of Distinct Signals As distance decreases, discrimination of signals dependent on synaptic tuning Extension:: Extension: How does signal strength modulate resolution of highly similar smells? Disagreement exists whether signal strength increases resolution The Model: The Model For Highly Overlapped Signal: G-ampa: Excitatory conductance of synapses G-gaba: Inhibitory conductance of synapses Slide 31: Observed Trends : As strength of stimulus is increased, general patterns emerge Network Displays: Recruitment over time Higher signal discrimination Normal Stimulus Double Magnitude Stimulus Time Time Extension:: Extension: How does signal strength modulate resolution of highly similar smells? Disagreement exists whether signal strength increases resolution Model suggests positive correlation Slide 33: Conclusion : This simple olfactory model seems to suggest: Spatially coded network displays ability to discriminate odors Both signal strength and relative synaptic conductance affect odor resolution Slide 34: The Learning Phenomena of Shadowing and Blocking Observed in Our Model: Slide 35: Proposed Mechanism for Shadowing: Inhibitory synaptic footprint at position in which overlap with excitatory synaptic footprint of neighboring stimulated cells in maximized Slide 36: Future Direction: - Imagine a scenario where stimuli are presented in ‘puffs’ in which the duration and sequence of puffs are varied. Make predictions regarding the ability of the network to resolve and learn separate stimuli as a function of synaptic footprint dimensions. Extend the model to more accurately reflect the multiple populations of different cells in the olfactory system. Slide 37: Acknowledgments: Thanks MBI and Dr. Friedman MBI Staff – Chris Dr. Terman Dr. Smith Dr. Wright

Related presentations


Other presentations created by Veronica1

497 Mobile Computing
04. 02. 2008
0 views

497 Mobile Computing

CATALOGO PERFUMES
10. 01. 2008
0 views

CATALOGO PERFUMES

salvage
10. 01. 2008
0 views

salvage

labcon2003
10. 01. 2008
0 views

labcon2003

summer05
10. 01. 2008
0 views

summer05

Goldman talk Chile2006
11. 01. 2008
0 views

Goldman talk Chile2006

ling lect 32
13. 01. 2008
0 views

ling lect 32

RocksUnderMicro tcm4 285395
14. 01. 2008
0 views

RocksUnderMicro tcm4 285395

Holi
14. 01. 2008
0 views

Holi

Care Premature Infant March02
15. 01. 2008
0 views

Care Premature Infant March02

TropicalRainforestMB
15. 01. 2008
0 views

TropicalRainforestMB

englishgrammarthemat rix
16. 01. 2008
0 views

englishgrammarthemat rix

koracin presentation
16. 01. 2008
0 views

koracin presentation

flood
16. 01. 2008
0 views

flood

2006 11 28 1
20. 01. 2008
0 views

2006 11 28 1

structured chaos
24. 01. 2008
0 views

structured chaos

PHYS125 lt5 ET
24. 01. 2008
0 views

PHYS125 lt5 ET

lad 005
05. 02. 2008
0 views

lad 005

1 Brief History of Neuroscience
05. 02. 2008
0 views

1 Brief History of Neuroscience

FTP arkivet6 liss
06. 02. 2008
0 views

FTP arkivet6 liss

voting4
09. 01. 2008
0 views

voting4

Chapter 1
28. 01. 2008
0 views

Chapter 1

AristotleEthics
28. 01. 2008
0 views

AristotleEthics

Ancient Greek Philosophy
29. 01. 2008
0 views

Ancient Greek Philosophy

02Eficiencia
30. 01. 2008
0 views

02Eficiencia

g poole pedagogies
31. 01. 2008
0 views

g poole pedagogies

art121 2000 East Greek
04. 02. 2008
0 views

art121 2000 East Greek

OralHealthGrade4
06. 02. 2008
0 views

OralHealthGrade4

ChiefLeschi web2
13. 02. 2008
0 views

ChiefLeschi web2

Intro to networking
14. 02. 2008
0 views

Intro to networking

Movie Genres and Stars
18. 02. 2008
0 views

Movie Genres and Stars

03Lect19TallBld
10. 01. 2008
0 views

03Lect19TallBld

valentinesday
22. 02. 2008
0 views

valentinesday

hypertensionCTU
25. 02. 2008
0 views

hypertensionCTU

dawid hepler slides
25. 02. 2008
0 views

dawid hepler slides

Medications for ADHD3
27. 02. 2008
0 views

Medications for ADHD3

epa 101 2007 dewey
03. 03. 2008
0 views

epa 101 2007 dewey

20010621e
26. 01. 2008
0 views

20010621e

BeSafeBeProudExpo07
07. 03. 2008
0 views

BeSafeBeProudExpo07

INTERSTATEHistoryfor SchoolsII
12. 03. 2008
0 views

INTERSTATEHistoryfor SchoolsII

13 Boylan GASIPmodeling
20. 02. 2008
0 views

13 Boylan GASIPmodeling

NEAD
14. 01. 2008
0 views

NEAD

20020928 ICFA HN
25. 03. 2008
0 views

20020928 ICFA HN

MODULE3D
19. 01. 2008
0 views

MODULE3D

Attracting Retail to Oakland
04. 02. 2008
0 views

Attracting Retail to Oakland

Tourism Trends of Philippines
28. 03. 2008
0 views

Tourism Trends of Philippines

medemerg
18. 01. 2008
0 views

medemerg

sf valentin
14. 02. 2008
0 views

sf valentin

111107 Angels Watching Over Me 2
11. 02. 2008
0 views

111107 Angels Watching Over Me 2

chuyw global culture text
19. 03. 2008
0 views

chuyw global culture text

internationaltransit ion
23. 01. 2008
0 views

internationaltransit ion

Loschen SuperBowlSurveillance 1
16. 04. 2008
0 views

Loschen SuperBowlSurveillance 1

smarter choices
05. 02. 2008
0 views

smarter choices

prezent Atameken eng
17. 01. 2008
0 views

prezent Atameken eng

bowmandicksonhandout 06
17. 01. 2008
0 views

bowmandicksonhandout 06

OhioStateGreatMoves
11. 01. 2008
0 views

OhioStateGreatMoves

baker1
05. 03. 2008
0 views

baker1

career resources
24. 03. 2008
0 views

career resources

Sensonor SW4 intro
11. 01. 2008
0 views

Sensonor SW4 intro

Faith Filled Living
15. 01. 2008
0 views

Faith Filled Living

luiz ramos
23. 01. 2008
0 views

luiz ramos

KathyrnLaBarre
29. 01. 2008
0 views

KathyrnLaBarre

Lu SF Presentation
21. 03. 2008
0 views

Lu SF Presentation

Presentation Noguera
14. 01. 2008
0 views

Presentation Noguera