Scribd
Upload
76 views36 pages

How Does A Neuron(s) Convey Information?

Neural codes convey information about stimuli via the firing patterns of neurons. Understanding neural codes is important for understanding brain function, diagnosing disorders, developing prosthetics and brain-computer interfaces. Neurons can encode information via single-neuron or population codes, and via rate codes or temporal codes. Rate codes represent information in average firing rates, while temporal codes represent information in precise spike timing. Population codes distribute information across coordinated or independent neural ensembles. Synchronized firing between neurons representing the same feature may help solve the binding problem of representing which features are associated. Overall, neural codes are an active area of research aimed at understanding how the brain represents and processes information.

Uploaded by

razorviper
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPT, PDF, TXT or read online on Scribd
76 views36 pages

How Does A Neuron(s) Convey Information?

Neural codes convey information about stimuli via the firing patterns of neurons. Understanding neural codes is important for understanding brain function, diagnosing disorders, developing prosthetics and brain-computer interfaces. Neurons can encode information via single-neuron or population codes, and via rate codes or temporal codes. Rate codes represent information in average firing rates, while temporal codes represent information in precise spike timing. Population codes distribute information across coordinated or independent neural ensembles. Synchronized firing between neurons representing the same feature may help solve the binding problem of representing which features are associated. Overall, neural codes are an active area of research aimed at understanding how the brain represents and processes information.

Uploaded by

razorviper
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPT, PDF, TXT or read online on Scribd
You are on page 1/ 36

How does a neuron(s)

convey information?

Neural Code

A code is a system of rules and mechanisms


by which a signal carries information
What is the neural code?

Neural Code Why bother?

To understand brain processes


Diagnose disorders
Prosthetic limbs
BCI brain-computer interface
etc.

Neural Code
Hypotheses
Single Neuron Coding
The Rate-Coding Hypothesis
The Temporal-Coding Hypothesis

Population Coding
The Independent-Coding Hypothesis
The Coordinated-Coding Hypothesis

Rate codes (averaging over time)


Simplest and oldest hypothesis (Adrian 1926, 1928)
The timing of the individual spikes is irrelevant, what
matters is the average rate. The rate hypothesis has
underpinned the vast majority of work.
Simple: we can
record neuron
present stimulus
Record the number of spikes in a certain period (100400 msecs)

The Rate-Coding Hypothesis


Only important aspect of spike train is mean firing rate.

Rate-Coding Tuning Curve

Rate-Coding Tuning Curve

Response of a V1 cell. Firing rate depends on the location and


orientation of the stimulus pattern. No firing to horizontal bars.
Diagonal bars generate more. 45 degree stimuli generate most.
Tuning curves may be narrowly or broadly tuned.

Rate-Coding - Value

Response of a cell in V1 depends on the contrast of the test


light. Increasing the contrast results in more action potentials.

Evidence for the rate hypothesis


Strategy: show that if we decode neurons using only rate
information, we get estimates of system performance
similar to the animals.
For instance:
For a motion detection experiment: For half of the neurons
in our study, the neurometric function derived from singleunit data was statistically indistinguishable from the
psychometric function measured on the same set of trials.
Britten et al (1992)
Similar results for contrast detection (Hawken and Parker
1990)

The firing rates of sensory nerves


encode the stimulus magnitude

Some alternatives to the rate


hypothesis
Temporal modulation of the spike train conveys information
(Optican, Richmond)
The first few spikes and their timing is all that matters
(Thorpe)
The exact time of each spike of a cell relative to that of other
cells is informative (Singer)
Every spike counts (Bialek)

Temporal-Coding
coding temporal structure in stimulus
The temporal structure of a spike-train carries information
about the temporal structure of the stimulus

Response of a neuron in MT in macaque to a (a) constant


velocity stimulus; (b) changing velocity stimulus

Temporal-Coding
coding spatial structure in stimulus
There is also evidence for temporal modulation of the
neuronal response to a static stimulus.
A variant of this type of temporal code might represent
information as the time at which spikes are produced
relative to the onset of a stimulus.
Richmond and Optican's (1990a, 1990b) measurements of
single neuron responses in the cortex of the monkey
suggest that the elapsed time since the onset of the
stimulus affects the encoding of visual form. Neurons
respond to a one-second presentation of a complex spatial
pattern with a time-varying spike rate.
The temporal details of the response depend on which
shape in the ensemble is presented.

Optican and Richmonds experiment


Showed a large number of simple, gridlike stimuli to a
macaque monkey
Looked at the firing rate patterns of the cells when the
stimuli were shown
Characterized these responses using principal components
analysis
Tried to see how much structure in the spike trains was
useful in guessing which stimulus was shown

Temporal modulation of response

Elapsed time since stimulus


onset affects the encoding of
visual form. The temporal
details of the cells response
depend on the shape in the
ensemble being presented.

Richmond & Optican, 1990a, 1990b

Temporal modulation of response


Spike-count (f0) is not
enough to discriminate
stimuli in the set, temporal
waveform (f1) helps.

f0

f1

Richmond & Optican, 1990a, 1990b

Temporal modulation of response


Except for a few stimuli,
the temporal waveform
code always transmits
more information than the
spike-count code

Dashed (solid) line is amount of


information transmitted per
pattern by the spike-count
(temporal waveform) code
Richmond & Optican, 1990a, 1990b

Temporal-Coding: Only the first

(few) spikes and latencies


matter

Neural coding using spike-firing times

The important signal for coding is the timing of the first spike of a cell
relative to that in other cells. More generally, the cells output is coded
by the rank of its spikes timing relative to all other cells
Van Rullen et al., 2005

Latency of the first spike


Very high information capacity

Invariant to illumination change: changing the contrast may change


Fast!!!!!!!!!
To calculate a rate, we need to wait until a fair number of spike
have been sent
To calculate the time to the first spike, we only need to wait for the
first spike

Behaviorally, we are indeed that fast!!!


We can respond very fast to visual stimuli (animal non-animal
distinctions can be made in photographs presented for only 80ms).
Rate codes will have trouble at these speeds. But some latency based
code will be fast enough.

Population Coding
Information is carried by pools of neurons
The Independent-Coding Hypothesis
each neuron contributes to the pool independently
the vote of each neuron gives a population vector

The Coordinated-Coding Hypothesis


the relationships among the neurons in a population is an
important part of the signal.
the signal cannot be decoded without considering spike
synchrony, oscillations, or some other relationship among
the neurons in the population.

Independent Coding - Value

Increases in the stimulus cause increases in the response of the


entire population. And if neurons respond to different ranges, the
population can discriminate a wide range of changes.

Independent Coding Tuning Curves

Changes in stimulus change the population that responds


optimally, and so the population vote can discriminate a wide
range of changes.

Coordinated Coding - Phase

The neurons fire at different phases with respect to the background


oscillation. The phase could code relevant information.

Coordinated Coding
phase-locked to a master clock

http://www.bio.brandeis.edu/lismanlab/research/figburst.jpg

Coordinated Coding - Synchrony

The upper four neurons are nearly synchronous, two other


neurons at the bottom are not synchronized with the others.
Synchronized neural activity may explain binding

Synchrony among neurons


cross-correlation
Often we want know not only which features are present in
the world, but which features belong together (the feature
binding problem)
One attractive proposal that has had a huge impact is that
cells that correspond to the same object fire together
This firing together can be of two related types:
Synchronous and oscillating (50 hertz)
Simply synchronous
Originally proposed to be happening by von der Malsburg
Experimental evidence for this was provided by Singer et
al.

The binding problem: why it arises

Synchrony among neurons


cross-correlation
Often we want to know not only which features are present
in the world, but which features belong together (the
feature binding problem)
One attractive proposal that has had a huge impact is that
cells that correspond to the same object fire together
This firing together can be of two related types:
Synchronous and oscillating (50 hertz)
Simply synchronous
Originally proposed by von der Malsburg
Experimental evidence for this was provided by Singer et
al.

Synchrony (cross-correlation)
Crosscorrelogram

PSTH

Crosscorrelogram

PSTH

Stimulus-dependent
synchronization in area
MT of awake macaque:
same motion direction
enhances synchrony in
spike firing, different
directions reduces
synchrony
Singer & Gray, 1995

Synchrony (cross-correlation)
Active neuronal populations can become organized into two assemblies
that are distinguished by the temporal coherence of activity within and
the lack of coherence between the cell groups responding to the two
different stimuli. Cells representing the same stimulus exhibit
synchronized response epochs, while no consistent correlations
occur between the responses of cells that are activated by different
stimuli.

Singer & Gray, 1995

Synchrony (cross-correlation)

The pattern of temporal


correlation among the
responses of individual
cells thus provides
additional information
regarding image
segmentation and
binding
Singer & Gray, 1995

Bottom line
There is ample evidence that, at least under some
circumstances, the timing of spikes cannot be
accounted for by a simple rate code.
There is potentially useful 'extra' information
contained in the timing of the first spike, or in the
spike trains of a single neuron or a population of
neurons.

Note of caution:
It is not been shown that this extra information is
actually used by neurons downstream.

Review
What is a neural code? Why is knowing about it
important?
Single neurons vs. population codes
Rate vs. temporal codes
Tuning curves.
Evidence for rate code
Criticisms of rate code and extensions to code that address
these criticisms.
Temporal waveform code. PCA? Information theory.
Integrate and fire neurons. Extension to first spike coding
What is binding problem? Why does it arise? Candidate
solution?
Synchrony. Singers statement about synchrony.
Criticisms of each candidate code.

You might also like