Nature Communications,
Journal Year:
2022,
Volume and Issue:
13(1)
Published: Jan. 10, 2022
There
are
two
principle
approaches
for
learning
in
artificial
intelligence:
error-driven
global
and
neuroscience-oriented
local
learning.
Integrating
them
into
one
network
may
provide
complementary
capabilities
versatile
scenarios.
At
the
same
time,
neuromorphic
computing
holds
great
promise,
but
still
needs
plenty
of
useful
algorithms
algorithm-hardware
co-designs
to
fully
exploit
its
advantages.
Here,
we
present
a
global-local
synergic
model
by
introducing
brain-inspired
meta-learning
paradigm
differentiable
spiking
incorporating
neuronal
dynamics
synaptic
plasticity.
It
can
meta-learn
plasticity
receive
top-down
supervision
information
multiscale
We
demonstrate
advantages
this
multiple
different
tasks,
including
few-shot
learning,
continual
fault-tolerance
vision
sensors.
achieves
significantly
higher
performance
than
single-learning
methods.
further
implement
Tianjic
platform
exploiting
prove
that
utilize
many-core
architecture
develop
hybrid
computation
paradigm.
Frontiers in Neural Circuits,
Journal Year:
2018,
Volume and Issue:
12
Published: July 31, 2018
Most
elementary
behaviors
such
as
moving
the
arm
to
grasp
an
object
or
walking
into
next
room
explore
a
museum
evolve
on
time
scale
of
seconds;
in
contrast,
neuronal
action
potentials
occur
few
milliseconds.
Learning
rules
brain
must
therefore
bridge
gap
between
these
two
different
scales.
Modern
theories
synaptic
plasticity
have
postulated
that
co-activation
pre-
and
postsynaptic
neurons
sets
flag
at
synapse,
called
eligibility
trace,
leads
weight
change
only
if
additional
factor
is
present
while
set.
This
third
factor,
signaling
reward,
punishment,
surprise,
novelty,
could
be
implemented
by
phasic
activity
neuromodulators
specific
inputs
special
events.
While
theoretical
framework
has
been
developed
over
last
decades,
experimental
evidence
support
traces
seconds
collected
during
years.
Here
we
review,
context
three-factor
plasticity,
four
key
experiments
role
combination
with
biological
implementation
neoHebbian
learning
rules.
Frontiers in Aging Neuroscience,
Journal Year:
2017,
Volume and Issue:
9
Published: Jan. 23, 2017
Motivation
can
have
invigorating
effects
on
behavior
via
dopaminergic
neuromodulation.
While
this
relationship
has
mainly
been
established
in
theoretical
models
and
studies
younger
subjects,
the
impact
of
structural
declines
system
during
healthy
aging
remains
unclear.
To
investigate
issue,
we
used
EEG
young
elderly
humans
a
reward-learning
paradigm.
Specifically,
scene
images
were
initially
encoded
by
combining
them
with
cues
predicting
monetary
reward
(high
vs.
low
reward).
Subsequently,
recognition
memory
for
scenes
was
tested.
As
main
finding,
show
that
response
times
encoding
faster
high
but
not
participants.
This
pattern
resembled
power
changes
theta-band
(4-7
Hz).
Importantly,
analyses
MRI
data
revealed
individual
reward-related
differences
elderlies'
time
could
be
predicted
integrity
substantia
nigra
(as
measured
magnetization
transfer).
These
findings
suggest
close
between
reward-based
invigoration,
theta
oscillations
age-dependent
system.
Trends in Neurosciences,
Journal Year:
2018,
Volume and Issue:
42(2), P. 102 - 114
Published: Nov. 16, 2018
Adaptation
to
the
ever-changing
world
is
critical
for
survival,
and
our
brains
are
particularly
tuned
remember
events
that
differ
from
previous
experiences.
Novel
experiences
induce
dopamine
release
in
hippocampus,
a
process
which
promotes
memory
persistence.
While
axons
ventral
tegmental
area
(VTA)
were
generally
thought
be
exclusive
source
of
hippocampal
dopamine,
recent
studies
have
demonstrated
noradrenergic
neurons
locus
coeruleus
(LC)
corelease
noradrenaline
hippocampus
their
boosts
retention
as
well.
In
this
opinion
article,
we
propose
projections
originating
VTA
LC
belong
two
distinct
systems
enhance
novel
events.
share
some
commonality
with
past
ones
('common
novelty')
activate
promote
semantic
formation
via
consolidation.
By
contrast,
bear
only
minimal
relationship
('distinct
trigger
strong
initial
consolidation
resulting
vivid
long-lasting
episodic
memories.
Cells,
Journal Year:
2021,
Volume and Issue:
10(4), P. 735 - 735
Published: March 26, 2021
Dopamine
(DA)
is
a
key
neurotransmitter
involved
in
multiple
physiological
functions
including
motor
control,
modulation
of
affective
and
emotional
states,
reward
mechanisms,
reinforcement
behavior,
selected
higher
cognitive
functions.
Dysfunction
dopaminergic
transmission
recognized
as
core
alteration
several
devastating
neurological
psychiatric
disorders,
Parkinson’s
disease
(PD),
schizophrenia,
bipolar
disorder,
attention
deficit
hyperactivity
disorder
(ADHD)
addiction.
Here
we
will
discuss
the
current
insights
on
role
DA
control
learning
mechanisms
its
involvement
synaptic
dynamics
through
different
pathways.
In
particular,
consider
neuromodulator
two
forms
plasticity,
known
long-term
potentiation
(LTP)
depression
(LTD)
cortical
subcortical
areas.
Finally,
delineate
how
effect
dendritic
spines
places
this
molecule
at
interface
between
systems.
Specifically,
be
focusing
PD,
vascular
dementia,
schizophrenia.
Trends in Cognitive Sciences,
Journal Year:
2019,
Volume and Issue:
23(12), P. 1014 - 1025
Published: Nov. 7, 2019
Curiosity
plays
a
fundamental
role
for
learning
and
memory,
but
the
neural
mechanisms
that
stimulate
curiosity
its
effect
on
memory
are
poorly
understood.
Accumulating
evidence
suggests
states
related
to
modulations
in
activity
dopaminergic
circuit
these
impact
encoding
consolidation
both
targets
of
incidental
information
encountered
during
states.
To
account
this
evidence,
we
propose
Prediction,
Appraisal,
Curiosity,
Exploration
(PACE)
framework,
which
attempts
explain
terms
cognitive
processes,
circuits,
behavior,
subjective
experience.
The
PACE
framework
generates
testable
predictions
can
future
investigation
underlying
curiosity-related
enhancements.
