Download Introduction to Artificial Neural Networks (ANNs)

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Connectome wikipedia , lookup

Computational creativity wikipedia , lookup

End-plate potential wikipedia , lookup

Multielectrode array wikipedia , lookup

Axon wikipedia , lookup

Neural modeling fields wikipedia , lookup

Neuroesthetics wikipedia , lookup

Single-unit recording wikipedia , lookup

Cortical cooling wikipedia , lookup

Artificial consciousness wikipedia , lookup

Clinical neurochemistry wikipedia , lookup

Neurotransmitter wikipedia , lookup

Neuroanatomy wikipedia , lookup

Binding problem wikipedia , lookup

Neurocomputational speech processing wikipedia , lookup

Neuroethology wikipedia , lookup

Neuroeconomics wikipedia , lookup

Stimulus (physiology) wikipedia , lookup

Nonsynaptic plasticity wikipedia , lookup

Optogenetics wikipedia , lookup

Holonomic brain theory wikipedia , lookup

Chemical synapse wikipedia , lookup

Neural oscillation wikipedia , lookup

Neural correlates of consciousness wikipedia , lookup

Synaptogenesis wikipedia , lookup

Catastrophic interference wikipedia , lookup

Molecular neuroscience wikipedia , lookup

Neural coding wikipedia , lookup

Central pattern generator wikipedia , lookup

Synaptic gating wikipedia , lookup

Artificial general intelligence wikipedia , lookup

History of artificial intelligence wikipedia , lookup

Channelrhodopsin wikipedia , lookup

Biological neuron model wikipedia , lookup

Artificial intelligence wikipedia , lookup

Neuropsychopharmacology wikipedia , lookup

Convolutional neural network wikipedia , lookup

Neural binding wikipedia , lookup

Metastability in the brain wikipedia , lookup

Neural engineering wikipedia , lookup

Development of the nervous system wikipedia , lookup

Artificial neural network wikipedia , lookup

Nervous system network models wikipedia , lookup

Recurrent neural network wikipedia , lookup

Types of artificial neural networks wikipedia , lookup

Transcript
Introduction to Artificial Neural Networks
(ANNs)
Keith L. Downing
The Norwegian University of Science and Technology (NTNU)
Trondheim, Norway
[email protected]
January 12, 2014
Keith L. Downing
Introduction to Artificial Neural Networks (ANNs)
NETtalk (Sejnowski + Rosenberg, 1986)
N
E
C
o
n
t
e
x
t
W
i
n
d
o
w
U
Letters
"Concepts"
Phonemes
R
O
S
Silent
C
I
E
N
C
E
IBM’s DECtalk: several man years of work → Reading machine.
NETtalk: 10 hours of backprop training on a 1000-word text, T1000.
95% accuracy on T1000; 78% accuracy on novel text.
Improvement during training sounds like a child learning to read.
Concept layer is key. 79 different (overlapping) clouds of neurons are
gradually formed, with each mapping to one of the 79 phonemes.
Keith L. Downing
Introduction to Artificial Neural Networks (ANNs)
Sample ANN Applications: Forecasting
1
Train the ANN (typically using backprop) on historical data to learn
[X (t−k ), X (t−k+1 ), . . . , X (t0 )] 7→ [X (t1 ), . . . , X (tm−1 ), X (tm )]
2
Use to predict future value(s) based on the past k values.
Sample applications (Ungar, in Handbook of Brain Theory and NNs, 2003)
Car sales
Airline passengers
Currency exchange rates
Electrical loads on regional power systems.
Flour prices
Stock prices (Warning: often tried, but few good, documented results).
Keith L. Downing
Introduction to Artificial Neural Networks (ANNs)
Brain-Computer Interfaces (BCI)
Scalp EEG
Neural
Ensembles
Neural
Context
1
2
3
Action
Ask subject to think about an activity (e.g. moving joystick left)
Register brain activity (EEG waves - non-invasive) or (Neural ensembles invasive)
ANN training case = (brain readings, joystick motion)
Sample applications (Millan, in Handbook of Brain Theory and NNs, 2003)
Keyboards (3 keystrokes per minute)
Artificial (prosthetic) hands
Wheelchairs
Computer games
Keith L. Downing
Introduction to Artificial Neural Networks (ANNs)
Brains as Bio-Inspiration
Texas
"Watermelon"
"The truth?
You can't handle
the truth."
"I got a 69
Chevy
with a 396..."
Grandmother
Distributed Memory - A key to the brain’s success, and a major
difference between it and computers.
Brain operations slower than computers, but massively parallel.
How can the brain inspire AI advances?
What is the proper level of abstraction?
Keith L. Downing
Introduction to Artificial Neural Networks (ANNs)
Signal Transmission in the Brain
AP
Axons
Nucleus
Dendrites
AP
Action Potential (AP)
A wave of voltage change along axons and dendrites.
Nucleus (soma) generates an AP if the sum of its incoming
APs is strong enough.
Keith L. Downing
Introduction to Artificial Neural Networks (ANNs)
Ion Channels
Repolarization
K+
K+
Ca++
Ca++
Na+
Na+
Depolarization
Keith L. Downing
Introduction to Artificial Neural Networks (ANNs)
Depolarization and Repolarization
Overshoot
+40 mV
Na+ gates close
K+ Efflux
K+ gates opens
K+ Efflux
0 mV
Na+ gates open
Na+ Influx
K+ gates close
-65 mV
Resting Potential
Time
Keith L. Downing
Undershoot
Introduction to Artificial Neural Networks (ANNs)
Transferring APs across a Synapse
Pre-synaptic
Terminal
Post-synaptic
Terminal
Synapse
NT-gated
Ion Channel
Vesicle
Action-Potential (AP)
Neurotransmitter (NT)
Neurotransmitters
Excite - Glutamate, AMPA ; bind Na+ and Ca++ channels.
Inhibit - GABA; binds K+ channels
Keith L. Downing
Introduction to Artificial Neural Networks (ANNs)
Location, Location, Location..of Synapses
I1
P2
Axons
Soma
Dendrites
P1
I2
Distal and Proximal Synapses
Synapses closer to the soma normally have a stronger effect.
Keith L. Downing
Introduction to Artificial Neural Networks (ANNs)
Donald Hebb (1949)
Fire Together, Wire Together
When an axon of cell A is near enough to excite a cell B and
repeatedly or persistently takes part in firing it, some growth
process or metabolic change takes place in one or both cells,
such that A’s efficiency as one of the cells firing B, is increased.
Hebb Rule
4wi,j = λ oi oj
Instrumental in Binding of..
pieces of an image
words of a song
multisensory input (e.g. words and images)
sensory inputs and proper motor outputs
simple movements of a complex action sequence
Keith L. Downing
Introduction to Artificial Neural Networks (ANNs)
Coincidence Detection and Synaptic Change
2 Key Synaptic Changes
1
the propensity to release neurotransmitter (and amount
released) at the pre-synaptic terminal,
2
the ease with which the post-synaptic terminal depolarizes
in the presence of neurotransmitters.
Coincidences
1
Pre-synaptic: Adenyl cyclase (AC) detects simultaneous
presence of Ca++ and serotonin.
2
Post-synaptic: NMDA receptors detect co-occurrence of
glutamate (a neurotransmitter) and depolarization.
Keith L. Downing
Introduction to Artificial Neural Networks (ANNs)
Pre-synaptic Modification
Depolarization
Salient Event
Ca++
Ca++
AC
5HT
ATP
cAMP
Pre-synaptic
Terminal
Post-synaptic
Terminal
PKA
Glutamate
NMDA Receptor
Mg++
AC
Adenyl
Cyclase
Keith L. Downing
5HT
Serotonin
Introduction to Artificial Neural Networks (ANNs)
Post-synaptic Modification
Polarized (relaxed)
postsynaptic
state
CA++
Net
Negative
Charge
Mg++
NMDA
Receptor
Depolarized (firing)
postsynaptic
state
Glutamate
CA++
Net
Positive
Charge
Mg++
Keith L. Downing
Introduction to Artificial Neural Networks (ANNs)
Neurochemical Basis of Hebbian Learning
Fire together: When the pre- and post-synaptic terminal
of a synapse depolarize at about the same time, the NMDA
channels on the post-synaptic side notice the coincidence
and open, thus allowing Ca++ to flow into the post-synaptic
terminal.
Wire together: Ca++ (via CaMKII and protein kinase C)
promotes post- and pre-synaptic changes that enhance the
efficiency of future AP transmission.
Keith L. Downing
Introduction to Artificial Neural Networks (ANNs)
Hebbian Basis of Classical Conditioning
Hear
Bell(CS)
S2
S1
Salivate
(R)
See
Food
(US)
Unconditioned Stimulus (US) - sensory input normally
associated with a response (R). E.g. the sight of food
stimulates salivation.
Conditioned Stimulus (CS) - sensory input having no
previous correlation with a response but which becomes
associated with it. E.g. Pavlov’s bell.
Keith L. Downing
Introduction to Artificial Neural Networks (ANNs)
Long-Term Potentiation (LTP)
Early Phase
Chemical changes to pre- and post-synaptic terminals, due to
AC and NMDA activity, respectively, increase the probability
(and efficiency) of AP transmission for minutes to hours after
training.
Late Phase
Structural changes occur to the link between the upstream
and downstream neuron. This often involves increases in the
numbers of axons and dendrites linking the two, and seems to
be driven by chemical processes triggered by high
concentrations of Ca++ in the post-synaptic soma.
Keith L. Downing
Introduction to Artificial Neural Networks (ANNs)
Abstraction
Human Brains
1011 neurons
1014 connections between them (a.k.a. synapses), many
modifiable
Complex physical and chemical activity to transmit ONE
action potential (AP) (a.k.a. signal) along ONE connection.
Artificial Neural Networks
N = 101 − 104 nodes
Max N 2 connections
All physics and chemistry represented by a few parameters
associated with nodes and arcs.
Keith L. Downing
Introduction to Artificial Neural Networks (ANNs)
Structural Abstraction
w
Node
w
w
w
Node
w
Node
Node
w
Node
Node
w
Soma
Axonal
Compartments
Dendritic
Compartments
Soma
Soma
Soma
Soma
Soma
Synapses
Dendrites
Soma
AP
Soma
Axons
AP
Soma
Keith L. Downing
Introduction to Artificial Neural Networks (ANNs)
Diverse ANN Topologies
A
D
B
E
Keith L. Downing
C
F
Introduction to Artificial Neural Networks (ANNs)
Functional Abstraction
Learn
N1
N2
w13
N3
Activate
Integrate
w12
Reset
RK
RNa
EK
ENa
VM
CM
Na+
K+
Lipid
bilayer
=
capacitor
Ion channel
=
resistor
K+
K+
Ca++
Ca++
Keith L. Downing
Na+
Na+
Introduction to Artificial Neural Networks (ANNs)
Main Functional Components
Learn
N1
N2
w13
Activate
Integrate
N3
w12
Reset
Integrate
neti = ∑nj=1 xj wi,j
Vi ← Vi + neti
:
Activate
xi =
1
1+e−Vi
Reset
Vi ← 0
Learn
4wi,j = λ xi xj
Keith L. Downing
Introduction to Artificial Neural Networks (ANNs)
Functional Options
Activate
Integrate
xi
Vi <= Vi + neti
Vi
i
xi
Vi
Reset
Never reset Vi
Spiking Neuron Model: Reset
Vi only when above threshold
Neurons without state:
Always reset Vi
Keith L. Downing
Introduction to Artificial Neural Networks (ANNs)
Activation Functions xi = f (Vi )
1
Step
Identity
1
xi
xi
0
0
Vi
Vi
T
Ramp
1
xi
0
Vi
T
Hyperbolic Tangent (tanh)
Logistic
1
1
xi
xi
-1
0
0
Vi
Keith L. Downing
0
Vi
Introduction to Artificial Neural Networks (ANNs)
Diverse Model Semantics
What Does xi Represent?
1
The occurrence of a spike in the action potential,
2
The instantaneous membrane potential of a neuron,
3
The firing rate of a neuron (AP’s / sec),
4
The average firing rate of a neuron over a time window,
5
The difference between a neuron’s current firing rate and
its average firing rate.
Keith L. Downing
Introduction to Artificial Neural Networks (ANNs)
Circuit Models of Neurons
Na+
K+
Ion channels
act as
resistors
Lipid
bilayer
acts
as a
capacitor
RK
RNa
EK
ENa
CM
Keith L. Downing
VM
Introduction to Artificial Neural Networks (ANNs)
Using Kirchoff’s Current Law
The sum of all currents into the cell must be zero.
The currents
i
: Capacitance: Icap = CM dV
dt
(VM −EK )
= gK (VM − EK )
rK
Na )
= gNa (VM − ENa )
(Sodium): INa = (VMr−E
Na
L)
= gL (VM − EL ) = Passive
(Leak): IL = (VMr−E
L
: Ionic (Potassium): IK =
: Ionic
: Ionic
ions through ungated channels.
flow of
where I = current, r = resistance, g = conductance ( 1r ), and VM
= membrane potential
Icap + IK + INa + IL = 0
CM
dVM
= −gK (VM − EK ) − gNa (VM − ENa ) − gL (VM − EL )
dt
Keith L. Downing
Introduction to Artificial Neural Networks (ANNs)
Modeling Voltage-Gated Channels
gK and gNa are sensitive to the membrane potential, VM
The gating probabilities
m, n and h = gating probabilities (between 0 and 1)
They are complex functions of VM , determined empirically
by Hodgkin and Huxley’s work on the giant squid axon.
Conductances are functions of the gating probabilities
gK = g K n4 - since 4 identical and independent parts of a K
gate need to be open.
g K = maximum K conductance.
gNa = g Na m3 h - since 3 identical and independent parts
(along with a different, 4th part) of an Na gate need to be
open.
g Na = maximum Na conductance.
Keith L. Downing
Introduction to Artificial Neural Networks (ANNs)
A Basic Version of the Hodgkin-Huxley Model
Repolarization
K+
K+
Ca++
Ca++
Na+
Na+
Depolarization
τm
dVM
= −gK (VM − EK ) − gNa (VM − ENa ) − gL (VM − EL )
dt
4VM ∝ Inflow(Na+) - outflow(K+) - Leak current
EL ≈ −60mV , EK ≈ −70mV , and ENa ≈ 50mV
τm includes the capacitance, CM .
Keith L. Downing
Introduction to Artificial Neural Networks (ANNs)
Leaky Integrate and Fire Neurons
xa
a
wia
i
b
xb
xc
c
wib
wic
xi
Vi
Leak
EL = -65 mV
These models ignore ion channels and activity along axons and dendrites.
Keith L. Downing
Introduction to Artificial Neural Networks (ANNs)
A Simple Leak-and-Integrate Model
τm
N
dVi
= cL (EL − Vi ) + cI ∑ xj wij
dt
j=1
(1)
Vi = intracellular potential for neuron i.
xi = output (current) from neuron i.
wij = weight on connection from j to i.
EL = extracellular potential
τm = membrane time const. Higher τm → slower change.
cL , cI = leak and integration constants.
A Common Abstraction
τm
N
dVi
= −Vi + ∑ xj wij
dt
j=1
Keith L. Downing
Introduction to Artificial Neural Networks (ANNs)
(2)
Firing Models
Continuous: Sigmoid Function
xi =
1
1 + e−cs Vi
(3)
* Often used for rate-coding, where xi = the neuron’s firing rate;
cs is a scaling constant.
Discrete: Step Function with Reset
1 if Vi > Tf
xi =
0 otherwise
Vi ← Vreset after exceeding the threshold, Tf .
Typical values: Vreset = −65mV , Tf = −50mV .
Often used in spiking neuron models, where xi is binary,
denoting presence or absence of an action potential.
Keith L. Downing
Introduction to Artificial Neural Networks (ANNs)
(4)
Temporal Abstraction
A
0.8
C
0.4
0.5
B
A
C
B
Time
+40 mV
0 mV
-65 mV
Time
Keith L. Downing
Introduction to Artificial Neural Networks (ANNs)
Spike Response Model (SRM) - Gerstner et. al., 2002
N
Vi (t) = κ(Iext ) + η(t − tˆi ) + ∑ wij
j=1
H
∑ εij (t − tˆi , t − tjh )
h=1
i
t*
!
ki
k
t*
j
!
kj
t*
The timing of each spike is very important in determining its effects upon
downstream neurons.
Keith L. Downing
Introduction to Artificial Neural Networks (ANNs)
Spiking Neurons
Eugene Izhikevich, 2003
A Simple Model of Spiking Neurons. IEEE Transactions on
Neural Networks, 14(6).
τm
N
dVi
= 0.04Vi2 + 5v + 140 − Ui + cI ∑ xj wij
dt
j=1
(5)
dUi
= a(bVi − Ui )
dt
(6)
τm
Ui = recovery factor
If Vi ≥ 30mV
then Vi ← Vreset , and Ui ← Ui + Ureset
Keith L. Downing
Introduction to Artificial Neural Networks (ANNs)
Parameterized Spiking Patterns
Regular Spiking
Chattering
Vi
Time
Intrinsic Bursting
Thalamocortical
Key parameters a, b, Vreset , and Ureset → spike patterns.
Keith L. Downing
Introduction to Artificial Neural Networks (ANNs)
Continuous Time Recurrent Neural Networks
Sensory Input Layer
1
2
3
4
5
Bias
Node
B
1
2
Hidden
Layer
1
2
Motor
Output
Layer
CTRNNs abstract away spikes but achieve complex dynamics with
neuron-specific time constants, gains and biases.
All weights evolve, but none are modified by learning.
Invented by Randall Beer in early 1990’s and used in many evolved,
minimally-cognitive agents.
Keith L. Downing
Introduction to Artificial Neural Networks (ANNs)
The Simple CTRNN Model
n
si =
∑ xj wi,j + Ii
j=1
dVi
1
= [−Vi + si + θi ]
dt
τi
1
xi =
1 + e−gi Vi
θi = bias; gi = gain.
τi = time constant for neuron i.
Each neuron implicitly runs at a different temporal resolution.
Keith L. Downing
Introduction to Artificial Neural Networks (ANNs)
Essence of Learning in Neural Networks
?
∆w
u1
w1
w2
u2
wn
v
post-synaptic neuron
un
pre-synaptic neurons
Most ANNs do not model spikes nor STDP. Learning is based
on a comparison of recent firing rates of neuron pairs.
Keith L. Downing
Introduction to Artificial Neural Networks (ANNs)
Spike-Timing Dependent Plasticity (STDP)
s
0.4
t
0
0
40 ms
-40 ms
-0.4
Change in synaptic strength (4s) as function of 4t = tpre − tpost ,
the times of the most recent pre- and post-synaptic spikes. The
maximum magnitude of change is roughly 0.4% of the
maximum possible synaptic strength/conductance.
Keith L. Downing
Introduction to Artificial Neural Networks (ANNs)
3 Fundamental ANN Learning Paradigms
Supervised
Constant, detailed feedback that includes the correct response
to each input; Omnipresent teacher.
Reinforced
Simple feedback mainly at the end of a problem-solving
attempt, although possibly a few intermediate rewards or
penalties, but no direct response recommendations.
Unsupervised
No feedback whatsoever. ANN normally tries to intelligently
cluster the inputs and/or learn proper correlations between
components of input space.
Keith L. Downing
Introduction to Artificial Neural Networks (ANNs)
Supervised Learning
You should have
turned RIGHT at
the last
intersection.
Sensory
Input
Correct
Action
Motor
Output
Error
∆W
Keith L. Downing
Introduction to Artificial Neural Networks (ANNs)
Reinforced Learning
You are at the goal!
Reinforcement
Signal
∆w
w
w
w
w
w
Keith L. Downing
Introduction to Artificial Neural Networks (ANNs)
Unsupervised Learning
A long trip down a
corridor is followed
by a left turn.
∆w
w
w
w
w
w
w
w
Keith L. Downing
Introduction to Artificial Neural Networks (ANNs)
Input
Hebbian Learning Rules
Basic Heterosynaptic
4wi = λ v (ui − θi )
Basic Homosynaptic
4wi = λ (v − θv )ui
General Hebbian
4wi = λ ui v
BCM
4wi = λ ui v (v − θv )
Oja
4wi = ui v − wi v 2
Homosynaptic
All active synapses are modified the same way, depending only on
the strength of the postsynaptic activity.
Heterosynaptic
Active synapses can be modified differently, depending upon the
strength of their presynaptic activity.
Keith L. Downing
Introduction to Artificial Neural Networks (ANNs)
Modelling Options to Consider
1
Single or multiple neurons?
2
Can neuron A send more than one axon to neuron B?
3
Are connections modeled as cables or just simple connector points (i.e.
a single weight).
4
Do neurons have state? I.e., does Vi (t + 1) depend on Vi (t)?
5
Do outputs (xi ) represent individual spikes or spike rates or ..?
6
Are neurons organized by layers?
7
Do layers follow a feed-forward topology or is there recurrence (i.e.
looping)?
8
Are neurons connected within layers or only between layers?
9
Is learning supervised, unsupervised or reinforced?
10
Is spike-timing dependent plasticity (STDP) involved in the learning
rule?
Keith L. Downing
Introduction to Artificial Neural Networks (ANNs)