retinal receptive fields Jonathan Schmok, Bochao Li, Yihan Zi - - PowerPoint PPT Presentation

β–Ά
retinal receptive fields
SMART_READER_LITE
LIVE PREVIEW

retinal receptive fields Jonathan Schmok, Bochao Li, Yihan Zi - - PowerPoint PPT Presentation

Model of motion-sensing retinal receptive fields Jonathan Schmok, Bochao Li, Yihan Zi Special thanks to Jun Wang The Retina - Directionally Sensitive(DS) Cells Why this project? Study of information coding in retina output firing pattern


slide-1
SLIDE 1

Model of motion-sensing retinal receptive fields

Jonathan Schmok, Bochao Li, Yihan Zi

Special thanks to Jun Wang

slide-2
SLIDE 2

The Retina

  • Directionally Sensitive(DS) Cells

Why this project? Study of information coding in retina output firing pattern

slide-3
SLIDE 3

Reichardt-Hassenstein Model

slide-4
SLIDE 4

RH Model with Neurons

Gap Junctions Chemical Synapse Leaky-Passive Neurons Hodgkin-Huxley Neuron

SYNAPSE MODEL NEURON MODEL

slide-5
SLIDE 5

The Network

rod 1 rod 0

𝜐 neuron 𝜐 neuron

arithmetic neuron arithmetic neuron

  • utput

neuron

slide-6
SLIDE 6

Input

Time Rod 1 Rod 0 Rod 1 Rod 0

slide-7
SLIDE 7

Rods

Input:

Parameter Value Capacitance 10 uF/cm^2 Leak Conductance 0.3 mS/cm^2 Resting Potential 0 mV

Model: Leaky Passive Neurons 𝑒𝑀 𝑒𝑒 = 𝐽𝑓𝑦𝑒 βˆ’ 𝑕𝑀 βˆ— π‘Š βˆ’ 𝐹𝑆 𝐷𝑛

slide-8
SLIDE 8

𝜐 Neurons

Input:

Parameter Value Capacitance 250 uF/cm^2 Leak Conductance 0.3 mS/cm^2 Resting Potential 0 mV

Model: Leaky Passive Neurons 𝑒𝑀 𝑒𝑒 = π½π‘•π‘π‘ž βˆ’ 𝑕𝑀 βˆ— π‘Š βˆ’ 𝐹𝑆 𝐷𝑛

  • Time delay from large capacitance
  • Representing high surface area starburst amacrine cells
slide-9
SLIDE 9

Arithmetic Neurons

Parameter Value Capacitance 10 uF/cm^2 Leak Conductance 0.3 mS/cm^2 Resting Potential 0 mV

Model: Leaky Passive Neurons 𝑒𝑀 𝑒𝑒 = (π½π‘•π‘π‘ž1βˆ— π½π‘•π‘π‘ž2) βˆ’ 𝑕𝑀 βˆ— π‘Š βˆ’ 𝐹𝑆 𝐷𝑛

  • Gap junction from input and tau neurons
  • Multiplication in neurons active area of

research

slide-10
SLIDE 10

Output Neuron

Parameter Value Capacitance 10 uF/cm^2 𝑕𝑀 0.3 ms/cm^2 𝐹𝑀

  • 65 mV

𝑕𝑂𝑏 100 mS/cm^2 𝑕𝐿 30 mS/cm^2 𝐹𝑂𝑏 40 mV 𝐹𝐿

  • 90 mV

Model: Hodgkin-Huxley Neuron 𝑒𝑀 𝑒𝑒 = 𝑕𝑀 βˆ— 𝐹𝑀 βˆ’ π‘Š + 𝐽𝐹𝑦𝑑𝑗𝑒𝑏𝑒𝑝𝑠𝑧 βˆ’ π½π½π‘œβ„Žπ‘—π‘π‘—π‘’π‘π‘ π‘§ βˆ’ 𝑕𝑂𝑏𝑛3β„Ž π‘Š βˆ’ 𝐹𝑂𝑏 βˆ’ π‘•πΏπ‘œ4 π‘Š βˆ’ 𝐹𝐿 𝐷𝑛

  • Excitatory synapse from left arithmetic unit

and inhibitory synapse from right arithmetic unit define preferred and null direction

  • 𝐹𝑀 determined manually for optimal

discrimination – other parameters from physiological measurements

slide-11
SLIDE 11

Reichardt-Hassenstain model

  • utputs of DS cells

Preferred Direction Null Direction

slide-12
SLIDE 12

Neural RH model of DS cells

Preferred Direction Null Direction

  • DS cells fire at their preferred direction and depolarize at null direction
  • Firing rate slowing down match natural neurons response stronger at the edge of signal
slide-13
SLIDE 13

Model analysis- input amplitude contrast

0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 0.3 0.5 1 2.5 5 7.5 10 12.5 15

Spke interval input contrast

Period=500

slide-14
SLIDE 14

Model analysis- velocity

Contrast for Spike Spike interval responds to velocity

0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.3 0.5 1 2.5 5 7.5 10 12.5 15

Spike interval input contrast

Period=200

0.01 0.02 0.03 0.04 0.05 0.06 0.3 0.5 1 2.5 5 7.5 10 12.5 15

SPike interval input contrast

Period=50

0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 0.3 0.5 1 2.5 5 7.5 10 12.5 15

Spke interval input contrast

Period=500

slide-15
SLIDE 15

Future Studies

  • How do neurons perform multiplication?
  • Design a model that is directionally sensitive in 2+ dimensions
  • Retina inspired computer vision
  • More complex neuronal models with spatial geometry