Modulation Channels in
Biomimic Artificial
Neurons
Richard B. Wells
1
The Biomimic Concept
• Purpose: Lay Out the General Architecture for
Biomimic Neurons
• Biomimic is a Subset of Biomimetic
– Biomimic = Mimicking the Neuron or Glial Cell
• The Artificial Neuron as a State Machine
– Classical Artificial Neurons After Training Are
“Combinational Analog Logic”
– Biological Neurons Are Better Modeled as Analog State
Machines
• “Information relates not so much to what you do
say as to what you could say
– Warren Weaver, 1964
2
Short-Term vs. Long-Term
Adaptation
• Conventional Adaptation = LTP, LTD
• Modulation = Short-Term Adaptation
• Short-Term Modulations => The Neuron as
a State Machine
• “In order to understand the style of
computation practiced by nervous systems,
we must study their hardware.”
– Christof Koch, 1999
3
Three Basic Modulation
Mechanisms in Biological
Neurons
• Voltage-Gated Ionic Channels (VGCs)
– Classical Hodgkin-Huxley Model
– Generally NOT “Inputs” (NMDA = ?)
• Ionotropic Channels -Synaptic Signal Inputs
• Metabotropic Receptors – “Control Inputs”
– “Intrinsic” vs. “Extrinsic” Information?
4
Voltage-Gated Channels
• VGCs Most Directly Control the Action
Potential
• They Are “Feedback Circuits” Mediated
Through the Integration of Membrane
Potential
• They Occur in Great Variety but Follow A
Generalized Architectural Scheme
5
Inactivating & Non-inactivating
VGCs
GV , t   G m p h
GV , t   G m
6
p
Principal Classes of VGCs
7
Ionotropic Channels
• Principal Transmission Inputs of Neuron
• IC s Recruit VGCs Through PSPs
8
The IC Synapse is a State
Machine
• Basic Mode is a Non-Retriggerable
Monostable Multivibrator
• Tetanic Facilitation = Input-Dependent
Synaptic Weight
• Delay, Conduction Time Can Be Modulated
(Phosphorylation)
9
Metabotropic Receptor Processes
• Their Role is Entirely Modulatory
• 2nd Messenger Processes Essential for
Stabilization of Synaptic Weight Changes (LTP,
LTD)
• Indirect Effects on Ion Current Flow, Generation
of Retrograde Messengers, Modulation of IC
Neurotransmitter Actions
• They Have Traditionally Been Ignored in Artificial
Neurons
10
Comparison of MSMS and IC
11
Primary Systems & Effects
12
Putting It All Together
E I N T X 's
W E IG H T
CONTROL
IC
F S M 's
I I N T X 'S
CURRENT
CONTROL
Ca
CURRENT
Ca
CURRENT
CO M M AND
G ATE
M O D U L A T IO N
CURRENT
CONTROL
VGC
F S M 's
Ca
LEVEL
G ATE
M O D U L A T IO N
E M N T X 'S
MSM
F S M 's
AP
AP
G e n e r a t io n
W E IG H T
CONTROL
AP
M O D U L A T IO N
Ca
B u ffe r
Ca
W E IG H T
CONTROL
Vm
A n a lo g
I n t e g r a t io n
C o re
Thr
THRESHOLD
M O D U L A T IO N
T M O S 'S
I M N T X 'S
T M I S 'S
13
Thank You. Questions?
14