bioRxiv (Cold Spring Harbor Laboratory),
Год журнала:
2018,
Номер
unknown
Опубликована: Сен. 16, 2018
Hippocampal
neurons
fire
selectively
in
local
behavioral
contexts
such
as
the
position
an
environment
or
phase
of
a
task,
1-3
and
are
thought
to
form
cognitive
map
task-relevant
variables.
1,4,5
However,
their
activity
varies
over
repeated
conditions,
6
different
runs
through
same
trials.
Although
widely
observed
across
brain,
7-10
variability
is
not
well
understood,
could
reflect
noise
structure,
encoding
additional
information.
6,11-13
Here,
we
introduce
conceptual
model
explain
terms
underlying,
population-level
structure
single-trial
neural
activity.
To
test
this
model,
developed
novel
unsupervised
learning
algorithm
incorporating
temporal
dynamics,
order
characterize
population
trajectory
on
nonlinear
manifold—a
space
possible
network
states.
The
manifold’s
captures
correlations
between
relationships
states,
constraints
arising
from
underlying
architecture
inputs.
Using
measurements
time
but
no
information
about
exogenous
variables,
recovered
hippocampal
manifolds
during
spatial
non-spatial
tasks
rats.
Manifolds
were
low-dimensional
smoothly
encoded
task-related
contained
extra
dimension
reflecting
beyond
measured
Consistent
with
our
fired
function
overall
state,
fluctuations
trials
corresponded
variation
manifold.
In
particular,
allowed
system
take
trajectories
despite
conditions.
Furthermore,
temporarily
decouple
current
conditions
traverse
neighboring
manifold
points
corresponding
past,
future,
nearby
Our
results
suggest
that
trial-to-trial
hippocampus
structured,
may
operation
internal
processes.
well-suited
for
organizing
support
memory,
1,5,14
planning,
12,15,16
reinforcement
learning.
17,18
general,
approach
find
broader
use
probing
organization
computational
role
circuit
dynamics
other
brain
regions.
Annual Review of Neuroscience,
Год журнала:
2016,
Номер
39(1), С. 237 - 256
Опубликована: Май 5, 2016
Brain
function
involves
the
activity
of
neuronal
populations.
Much
recent
effort
has
been
devoted
to
measuring
populations
in
different
parts
brain
under
various
experimental
conditions.
Population
patterns
contain
rich
structure,
yet
many
studies
have
focused
on
pairwise
relationships
between
members
a
larger
population-termed
noise
correlations.
Here
we
review
progress
understanding
how
these
correlations
affect
population
information,
information
should
be
quantified,
and
what
mechanisms
may
give
rise
As
coding
theory
improved,
it
made
clear
that
some
forms
correlation
are
more
important
for
than
others.
We
argue
this
is
critical
lesson
those
interested
responses
generally:
Descriptions
motivated
by
linked
well-specified
function.
Within
context,
offer
suggestions
where
current
theoretical
frameworks
fall
short.
Annual Review of Neuroscience,
Год журнала:
2017,
Номер
40(1), С. 557 - 579
Опубликована: Июнь 9, 2017
Inhibitory
neurons,
although
relatively
few
in
number,
exert
powerful
control
over
brain
circuits.
They
stabilize
network
activity
the
face
of
strong
feedback
excitation
and
actively
engage
computations.
Recent
studies
reveal
importance
a
precise
balance
inhibition
neural
circuits,
which
often
requires
exquisite
fine-tuning
inhibitory
connections.
We
review
synaptic
plasticity
its
roles
shaping
both
feedforward
control.
discuss
necessity
complex,
codependent
mechanisms
to
build
nontrivial,
functioning
networks,
we
end
by
summarizing
experimental
evidence
such
interactions.
Science,
Год журнала:
2016,
Номер
352(6291), С. 1319 - 1322
Опубликована: Июнь 9, 2016
Monocular
deprivation
evokes
a
prominent
shift
of
neuronal
responses
in
the
visual
cortex
toward
open
eye,
accompanied
by
functional
and
structural
synaptic
rearrangements.
This
is
reversible,
but
it
unknown
whether
recovery
happens
at
level
individual
neurons
or
reflects
population
effect.
We
used
ratiometric
Ca(2+)
imaging
to
follow
activity
same
excitatory
layer
2/3
mouse
over
months
during
repeated
episodes
ocular
dominance
(OD)
plasticity.
observed
robust
shifts
eye
most
neurons.
Nevertheless,
these
cells
faithfully
returned
their
pre-deprivation
OD
binocular
recovery.
Moreover,
initial
network
correlation
structure
was
largely
recovered,
suggesting
that
connectivity
may
be
regained
despite
experience-dependent
Neuron,
Год журнала:
2018,
Номер
98(4), С. 846 - 860.e5
Опубликована: Май 1, 2018
Correlated
variability
in
cortical
activity
is
ubiquitously
quenched
following
stimulus
onset,
a
stimulus-dependent
manner.
These
modulations
have
been
attributed
to
circuit
dynamics
involving
either
multiple
stable
states
("attractors")
or
chaotic
activity.
Here
we
show
that
qualitatively
different
dynamical
regime,
fluctuations
about
single,
stimulus-driven
attractor
loosely
balanced
excitatory-inhibitory
network
(the
stochastic
"stabilized
supralinear
network"),
best
explains
these
modulations.
Given
the
input/output
functions
of
neurons,
increased
drive
strengthens
effective
connectivity.
This
shifts
balance
from
interactions
amplify
suppressive
inhibitory
feedback,
quenching
correlated
around
more
strongly
driven
steady
states.
Comparing
previously
published
and
original
data
analyses,
this
mechanism,
unlike
previous
proposals,
uniquely
accounts
for
spatial
patterns
fast
temporal
suppression.
Specifying
operating
regime
key
understanding
computations
underlying
perception.
