Journal of Neuroscience,
Год журнала:
2022,
Номер
42(45), С. 8460 - 8467
Опубликована: Ноя. 9, 2022
Dendrites
receive
the
vast
majority
of
a
single
neuron's
inputs,
and
coordinate
transformation
these
signals
into
neuronal
output.
Ex
vivo
theoretical
evidence
has
shown
that
dendrites
possess
powerful
processing
capabilities,
yet
little
is
known
about
how
mechanisms
are
engaged
in
intact
brain
or
they
influence
circuit
dynamics.
New
experimental
computational
technologies
have
led
to
surge
interest
unravel
harness
their
potential.
This
review
highlights
recent
emerging
work
combines
established
cutting-edge
identify
role
function.
We
discuss
active
dendritic
mediation
sensory
perception
learning
neocortical
hippocampal
pyramidal
neurons.
Complementing
physiological
findings,
we
present
provides
new
insights
underlying
computations
neurons
networks
by
using
biologically
plausible
implementations
processes.
Finally,
novel
brain-computer
interface
task,
which
assays
somatodendritic
coupling
study
biological
credit
assignment.
Together,
findings
exciting
progress
understanding
critical
for
behavior,
highlight
subcellular
processes
can
contribute
our
both
artificial
neural
computation.
Nature Communications,
Год журнала:
2024,
Номер
15(1)
Опубликована: Янв. 4, 2024
It
is
widely
believed
the
brain-inspired
spiking
neural
networks
have
capability
of
processing
temporal
information
owing
to
their
dynamic
attributes.
However,
how
understand
what
kind
mechanisms
contributing
learning
ability
and
exploit
rich
properties
satisfactorily
solve
complex
computing
tasks
in
practice
still
remains
be
explored.
In
this
article,
we
identify
importance
capturing
multi-timescale
components,
based
on
which
a
multi-compartment
model
with
dendritic
heterogeneity,
proposed.
The
enables
dynamics
by
automatically
heterogeneous
timing
factors
different
branches.
Two
breakthroughs
are
made
through
extensive
experiments:
working
mechanism
proposed
revealed
via
an
elaborated
XOR
problem
analyze
feature
integration
at
levels;
comprehensive
performance
benefits
over
ordinary
achieved
several
benchmarks
for
speech
recognition,
visual
electroencephalogram
signal
robot
place
shows
best-reported
accuracy
compactness,
promising
robustness
generalization,
high
execution
efficiency
neuromorphic
hardware.
This
work
moves
significant
step
toward
real-world
applications
appropriately
exploiting
biological
observations.
Back-propagating
action
potentials
(bAPs)
regulate
synaptic
plasticity
by
evoking
voltage-dependent
calcium
influx
throughout
dendrites.
Attenuation
of
bAP
amplitude
in
distal
dendritic
compartments
alters
a
location-specific
manner
reducing
bAP-dependent
influx.
However,
it
is
not
known
if
neurons
exhibit
branch-specific
variability
signals,
independent
distance-dependent
attenuation.
Here,
we
reveal
that
bAPs
fail
to
evoke
through
voltage-gated
channels
(VGCCs)
specific
population
branches
mouse
cortical
layer
2/3
pyramidal
cells,
despite
substantial
VGCC-mediated
sister
branches.
These
contain
VGCCs
and
successfully
propagate
the
absence
input;
nevertheless,
they
bAP-evoked
due
reduction
amplitude.
We
demonstrate
these
have
more
elaborate
branch
structure
compared
branches,
which
causes
local
electrotonic
impedance
Finally,
show
still
amplify
synaptically-mediated
because
differences
voltage-dependence
kinetics
NMDA-type
glutamate
receptors.
Branch-specific
compartmentalization
signals
may
provide
mechanism
for
diversify
tuning
across
tree.
Advanced Electronic Materials,
Год журнала:
2025,
Номер
unknown
Опубликована: Фев. 2, 2025
Abstract
Emerging
memristor
synapses
with
ion
dynamics
have
the
potential
to
process
spatiotemporal
information
and
can
accelerate
development
of
energy‐efficient
neuromorphic
computing.
However,
conventional
ion‐migration‐type
memristors
suffer
from
low
switching
speed
uncontrollable
conductance
modulation,
hindering
hardware
implementation.
Here,
intercalation‐mediated
in
MoS
2
is
introduced
for
a
highly
synapse
(HEMS)
accurately
emulate
bio‐synaptic
function.
Li‐ion
intercalation
into
few‐layer
induce
structural
evolution,
thereby
achieving
high‐speed
controllable
modulation
HEMS.
Consequently,
HEMS
exhibits
energy
efficiency
fast
500
ns
consumption
2.85
fJ
per
synaptic
event.
The
stable
bidirectional
plasticity
by
consecutive
voltage
pulses
5000
times
be
achieved
Besides,
endowed
logic
functions
multiple
sets
inputs
parallel
integration.
This
work
offers
an
alternative
strategy
fast‐speed
via
develop
future
Nature Communications,
Год журнала:
2023,
Номер
14(1)
Опубликована: Сен. 18, 2023
Abstract
Biophysically
detailed
multi-compartment
models
are
powerful
tools
to
explore
computational
principles
of
the
brain
and
also
serve
as
a
theoretical
framework
generate
algorithms
for
artificial
intelligence
(AI)
systems.
However,
expensive
cost
severely
limits
applications
in
both
neuroscience
AI
fields.
The
major
bottleneck
during
simulating
compartment
is
ability
simulator
solve
large
systems
linear
equations.
Here,
we
present
novel
D
endritic
H
ierarchical
S
cheduling
(DHS)
method
markedly
accelerate
such
process.
We
theoretically
prove
that
DHS
implementation
computationally
optimal
accurate.
This
GPU-based
performs
with
2-3
orders
magnitude
higher
speed
than
classic
serial
Hines
conventional
CPU
platform.
build
DeepDendrite
framework,
which
integrates
GPU
computing
engine
NEURON
demonstrate
tasks.
investigate
how
spatial
patterns
spine
inputs
affect
neuronal
excitability
human
pyramidal
neuron
model
25,000
spines.
Furthermore,
provide
brief
discussion
on
potential
AI,
specifically
highlighting
its
enable
efficient
training
biophysically
typical
image
classification
The Journal of Physiology,
Год журнала:
2022,
Номер
601(15), С. 3091 - 3102
Опубликована: Окт. 11, 2022
Abstract
For
the
past
seven
decades,
Hodgkin–Huxley
(HH)
formalism
has
been
an
invaluable
tool
in
arsenal
of
neuroscientists,
allowing
for
robust
and
reproducible
modelling
ionic
conductances
electrophysiological
phenomena
they
underlie.
Despite
its
apparent
age,
role
as
a
cornerstone
computational
neuroscience
not
waned.
The
discovery
dendritic
regenerative
events
mediated
by
synaptic
solidified
importance
HH‐based
models
further,
yielding
new
predictions
concerning
integration,
plasticity
neuronal
computation.
These
are
often
validated
through
vivo
vitro
experiments,
advancing
our
understanding
neuron
biological
system
emphasizing
detailed
instrument
research.
In
this
article,
we
discuss
recent
studies
which
HH
is
used
to
shed
light
on
function
phenomena.
image
Applied Physics Letters,
Год журнала:
2023,
Номер
123(1)
Опубликована: Июль 3, 2023
Artificial
leaky
integrate-and-fire
(LIF)
neurons
have
attracted
significant
attention
for
building
brain-like
computing
and
neuromorphic
systems.
However,
previous
artificial
LIF
primarily
focused
on
implementing
function,
the
function
of
dendritic
modulation
has
rarely
been
reported.
In
this
Letter,
a
tunable
neuron
based
an
IGZO
electric-double-layer
(EDL)
transistor
TaOx
memristor
is
fabricated,
investigated.
An
IGZO-based
EDL
with
modulatory
terminal
used
to
realize
nonlinear
integration
filtering
capability,
as
well
neural
excitability.
Ag/TaOx/ITO
threshold
switching
mimics
all-or-nothing
spiking
soma.
By
incorporating
these
two
components
in
customized
way,
such
can
emulate
key
biological
rich
computational
flexibility.
Our
dynamics
great
potential
perform
more
complex
tasks
future
Frontiers in Neuroscience,
Год журнала:
2024,
Номер
17
Опубликована: Янв. 31, 2024
The
unique
characteristics
of
neocortical
pyramidal
neurons
are
thought
to
be
crucial
for
many
aspects
information
processing
and
learning
in
the
brain.
Experimental
data
suggests
that
their
segregation
into
two
distinct
compartments,
basal
dendrites
close
soma
apical
branching
out
from
thick
dendritic
tuft,
plays
an
essential
role
cortical
organization.
A
recent
hypothesis
states
layer
5
cells
associate
top-down
contextual
arriving
at
tuft
with
features
sensory
input
predominantly
arrives
dendrites.
It
has
however
remained
unclear
whether
such
context
association
could
established
by
synaptic
plasticity
processes.
In
this
work,
we
formalize
objective
through
a
mathematical
loss
function
derive
rule
synapses
optimizes
loss.
resulting
utilizes
is
available
either
locally
synapse,
branch-local
NMDA
spikes,
or
global
Ca
2+
events,
both
which
have
been
observed
experimentally
cells.
We
show
computer
simulations
enables
patterns
high
somatic
activity.
Furthermore,
it
networks
neuron
models
perform
context-dependent
tasks
continual
allocating
new
branches
novel
contexts.