📝 Original Info
- Title: Discovering Patterns in Multi-neuronal Spike Trains using the Frequent Episode Method
- ArXiv ID: 0709.0566
- Date: 2008-03-10
- Authors: Researchers from original ArXiv paper
📝 Abstract
Discovering the 'Neural Code' from multi-neuronal spike trains is an important task in neuroscience. For such an analysis, it is important to unearth interesting regularities in the spiking patterns. In this report, we present an efficient method for automatically discovering synchrony, synfire chains, and more general sequences of neuronal firings. We use the Frequent Episode Discovery framework of Laxman, Sastry, and Unnikrishnan (2005), in which the episodes are represented and recognized using finite-state automata. Many aspects of functional connectivity between neuronal populations can be inferred from the episodes. We demonstrate these using simulated multi-neuronal data from a Poisson model. We also present a method to assess the statistical significance of the discovered episodes. Since the Temporal Data Mining (TDM) methods used in this report can analyze data from hundreds and potentially thousands of neurons, we argue that this framework is appropriate for discovering the `Neural Code'.
💡 Deep Analysis
Deep Dive into Discovering Patterns in Multi-neuronal Spike Trains using the Frequent Episode Method.
Discovering the ‘Neural Code’ from multi-neuronal spike trains is an important task in neuroscience. For such an analysis, it is important to unearth interesting regularities in the spiking patterns. In this report, we present an efficient method for automatically discovering synchrony, synfire chains, and more general sequences of neuronal firings. We use the Frequent Episode Discovery framework of Laxman, Sastry, and Unnikrishnan (2005), in which the episodes are represented and recognized using finite-state automata. Many aspects of functional connectivity between neuronal populations can be inferred from the episodes. We demonstrate these using simulated multi-neuronal data from a Poisson model. We also present a method to assess the statistical significance of the discovered episodes. Since the Temporal Data Mining (TDM) methods used in this report can analyze data from hundreds and potentially thousands of neurons, we argue that this framework is appropriate for discovering the `
📄 Full Content
arXiv:0709.0566v2 [cs.DB] 26 Sep 2007
Dis o
v
ering
P
atterns
in
Multi-neuronal
Spik
e
T
rains
using
the
F
requen
t
Episo
de
Metho
d
K.P
.Unnikrishnan∗
and
Debprak
ash
P
atnaik†
and
P
.S.Sastry‡
ABSTRA
CT
Dis o
v
ering
the
'Neural
Co
de'
from
m
ulti-neuronal
spik
e
trains
is
an
imp
or-
tan
t
task
in
neuros ien e.
F
or
su
h
an
analysis,
it
is
imp
ortan
t
to
unearth
in
teresting
regularities
in
the
spiking
patterns.
In
this
rep
ort,
w
e
presen
t
an
ef-
ien
t
metho
d
for
automati ally
dis o
v
ering
syn
hron
y
,
syn re
hains,
and
more
general
sequen es
of
neuronal
rings.
W
e
use
the
F
requen
t
Episo
de
Dis o
v
ery
framew
ork
of
Laxman,
Sastry
,
and
Unnikrishnan
(2005),
in
whi
h
the
episo
des
are
represen
ted
and
re ognized
using
nite-state
automata.
Man
y
asp
e ts
of
fun tional
onne tivit
y
b
et
w
een
neuronal
p
opulations
an
b
e
inferred
from
the
episo
des.
W
e
demonstrate
these
using
sim
ulated
m
ulti-neuronal
data
from
a
P
oisson
mo
del.
W
e
also
presen
t
a
metho
d
to
assess
the
statisti al
signi an e
of
the
dis o
v
ered
episo
des.
Sin e
the
T
emp
oral
Data
Mining
(TDM)
metho
ds
used
in
this
rep
ort
an
analyze
data
from
h
undreds
and
p
oten
tially
thousands
of
neurons,
w
e
argue
that
this
framew
ork
is
appropriate
for
dis o
v
ering
the
`Neural
Co
de’.
1
INTR
ODUCTION
Analyzing
spik
e
trains
from
h
undreds
of
neurons
is
an
imp
ortan
t
and
ex iting
problem.
By
using
exp
erimen
tal
te
hniques
su
h
as
Mi ro
Ele tro
de
Arra
ys
or
imaging
of
neural
urren
ts
through
v
oltage-sensitiv
e
dy
es
et .,
spik
e
data
an
b
e
re orded
sim
ultaneously
from
man
y
neurons
[1,
2℄.
Automati ally
dis o
v
ering
∗
General
Motors
R&D
Cen
ter,
W
arren,
MI
†
Dept.
Ele etri al
Engineering,
Indian
Institute
of
S ien e,
Bangalore
‡
Dept.
Ele etri al
Engineering,
Indian
Institute
of
S ien e,
Bangalore
patterns
(regularities)
in
these
spik
e
trains
an
lead
to
b
etter
understanding
of
the
fun tional
relationships
within
the
system
that
pro
du ed
the
spik
es.
Su
h
understanding
of
fun tional
relations
em
b
edded
in
spik
e
trains
lead
to
man
y
appli ations,
e.g.,
b
etter
brain-ma
hine
in
terfa es.
Su
h
an
analysis
an
also
ultimately
allo
w
us
to
systemati ally
answ
er
the
question,
“is
there
a
neural
o
de?”.
In
this
pap
er,
w
e
presen
t
some
no
v
el
metho
ds
to
analyze
spik
e
train
data,
based
on
the
metho
d
of
frequen
t
episo
de
dis o
v
ery
in
time-ordered
ev
en
t
se-
quen es
[3
,
4
,
5
℄,
whi
h
is
from
the
eld
of
temp
oral
data
mining.
T
emp
oral
data
mining
is
on erned
with
analysis
of
large
sequen
tial
data
sets
[6
℄.
Su
h
data
sets
with
temp
oral
dep
enden ies
frequen
tly
o
ur
in
man
y
business,
engi-
neering
and
s ien
ti
s enarios.
F
requen
t
episo
de
dis o
v
ery
,
originally
prop
osed
in
[3
℄,
is
one
of
the
p
opular
framew
orks
in
temp
oral
data
mining.
Here,
the
data
is
view
ed
as
a
time-ordered
sequen e
of
ev
en
ts
where
ea
h
ev
en
t
is
hara ter-
ized
b
y
an
ev
en
t
t
yp
e
and
a
time
of
o
urran e.
A
few
examples
of
su
h
data
are
alarms
in
a
tele omm
uni ation
net
w
ork,
fault
logs
of
a
man
ufa turing
plan
t
et .
The
goal
of
the
analysis
is
to
unearth
temp
oral
patterns
( alled
episo
des)
that
o
ur
su ien
tly
often
along
that
sequen e.
These
dis o
v
ered
patterns
are
alled
frequen
t
episo
des.
The
m
ulti-neuronal
spik
e
train
data
is
also
a
sequen
tial
or
time-ordered
data
stream
of
ev
en
ts
where
ea
h
ev
en
t
is
a
spik
e
at
a
parti u-
lar
time
and
the
ev
en
t
t
yp
e
w
ould
b
e
the
neuron
(or
the
ele tro
de
in
the
mi ro
ele tro
de
arra
y)
that
generated
the
spik
e.
Sin e
fun tionally
in
ter onne ted
neurons
tend
to
re
in
ertain
pre ise
patterns,
dis o
v
ering
frequen
t
patterns
in
su
h
temp
oral
data
an
help
understand
the
underlying
neural
ir uitry
.
In
this
pap
er,
w
e
argue
that
the
frequen
t
episo
des
framew
ork
is
ideally
suited
for
su
h
analysis.
There
are
e ien
t
algorithms
for
automati ally
dete ting
man
y
t
yp
es
of
frequen
t
episo
des
[3,
4
℄.
Ho
w
ev
er,
as
w
e
shall
see,
in
analyzing
neural
spiking
data,
one
needs
metho
ds
that
an
dis o
v
er
frequen
t
episo
des
under
dif-
feren
t
kinds
of
temp
oral
onstrain
ts.
W
e
explain
some
datamining
algorithms
for
frequen
t
episo
de
dis o
v
ery
under
su
h
temp
oral
onstrain
ts
[5
℄.
Through
extensiv
e
sim
ulation
studies
using
b
oth
syn
theti
and
real
neural
data,
w
e
ar-
gue
that
the
frequen
t
episo
des
framew
ork
is
ideally
suited
for
this
appli ation.
W
e
sho
w
that
these
datamining
te
hniques
pro
vide
a
v
ery
e ien
t
and
general
purp
ose
metho
dology
for
dete ting
man
y
t
yp
es
of
in
teresting
patterns
in
spik
e
2
data.
Most
of
the
urren
tly
a
v
ailable
metho
ds
for
analyzing
spik
e
train
data
rely
on
quan
tities
that
an
b
e
omputed
through
ross
orrelations
among
spik
e
trains
(time
shifted
with
resp
e t
to
one
another)
to
iden
tify
in
teresting
patterns
in
spiking
a tivit
y
.
There
are
metho
ds
to
lo
ok
for
sp
e i
patterns
and
assess
their
statisti al
signi an e
under
a
n
ull
h
yp
othesis
that
dieren
t
spik
e
trai
…(Full text truncated)…
Reference
This content is AI-processed based on ArXiv data.