High-Precision Attention Mechanism for Machine Vision Enabled by an Artificial Optoelectronic Memristor Synapse
Nano Letters,
Год журнала:
2025,
Номер
unknown
Опубликована: Фев. 5, 2025
The
rapid
advancement
of
artificial
intelligence
has
facilitated
the
broad
application
machine
vision
systems
in
diverse
industries.
However,
these
are
often
confronted
with
computational
challenges
stemming
from
an
overwhelming
amount
data.
Here,
we
have
developed
a
novel
optoelectronic
memristor
synapse
constructed
ITO/Nb:SrTiO3
heterostructure,
which
synergistically
integrates
light
signal
detection
information
processing
and
memory
functions.
Notably,
achieved
precise
decoupling
interactions
between
power
wavelength
at
hardware
level,
significantly
enhancing
accuracy
efficiency
image
processing.
Furthermore,
by
incorporating
attention
mechanism
analogous
to
that
human
vision,
enabled
device
weight
key
filter
out
irrelevant
Experimental
results
demonstrate
this
can
increase
facial
recognition
13%
while
reducing
data
load
35–65%.
This
work
is
expected
advance
development
synapses
vision.
Язык: Английский
Optoelectronic Neuromorphic Logic Memory Device Based on Ga2O3/MoS2 Van der Waals Heterostructure with High Rectification and On/Off Ratios
Advanced Functional Materials,
Год журнала:
2024,
Номер
unknown
Опубликована: Июль 30, 2024
Abstract
It
is
crucial
to
develop
advanced
optoelectronic
devices
that
incorporate
multiple
functions,
including
sensing,
storage,
and
computing,
which
considered
at
the
forefront
of
semiconductor
optoelectronics
meet
emerging
functional
diversification.
In
this
study,
by
stacking
n‐type
Ga
2
O
3
with
MoS
flakes,
a
/MoS
heterostructure
device
high
rectification
ratio
≈10
5
on/off
8
fabricated,
achieves
detectivity
1.34
×
10
9
Jones
responsivity
28.92
mA/W.
More
importantly,
shows
potential
ability
integrate
sensing
memorizing,
simultaneously,
can
be
used
as
artificial
neuromorphic
synaptic.
The
exhibits
excellent
photo‐induced
synaptic
functions
short‐term
plasticity,
long‐term
paired‐pulse
facilitation,
realizing
couple
light
electrical
signals
Pavlovian
associative
learning.
At
last,
also
demonstrates
information
processing
act
logic
gate
AND
synergistically
regulating
states
voltage.
research
introduces
an
innovative
strategy
for
development
next‐generation
are
highly
integrated
memory,
demonstrating
great
application
prospects
in
constructing
efficient
visual
systems.
Язык: Английский
Self‐Powered Artificial Neuron Devices: Towards the All‐In‐One Perception and Computation System
Advanced Materials,
Год журнала:
2025,
Номер
unknown
Опубликована: Фев. 18, 2025
Abstract
The
increasing
demand
for
energy
supply
in
sensing
units
and
the
computational
efficiency
of
computation
has
prompted
researchers
to
explore
novel,
integrated
technology
that
offers
high
low
consumption.
Self‐powered
enables
environmental
perception
without
external
sources,
while
neuromorphic
provides
energy‐efficient
high‐performance
computing
capabilities.
integration
self‐powered
presents
a
promising
solution
an
all‐in‐one
system.
This
review
examines
recent
developments
advancements
artificial
neuron
devices
based
on
triboelectric,
piezoelectric,
photoelectric
effects,
focusing
their
structures,
mechanisms,
functions.
Furthermore,
it
compares
electrical
characteristics
various
types
discusses
effective
methods
enhancing
performance.
Additionally,
this
comprehensive
summary
systems,
encompassing
tactile,
visual,
auditory
systems.
Moreover,
elucidates
recently
systems
combine
perception,
computing,
actuation
into
configurations,
aspiring
realize
closed‐loop
control.
seamless
holds
significant
potential
shaping
more
intelligent
future
humanity.
Язык: Английский
Neuroscience-Driven Solutions for Speech Impairment
Advances in medical education, research, and ethics (AMERE) book series,
Год журнала:
2025,
Номер
unknown, С. 119 - 146
Опубликована: Янв. 10, 2025
Speech
impairment
presents
a
significant
challenge
for
aged
and
chronically
ill
individuals.
Emerging
assistive
technologies
offer
promising
solutions
to
improve
their
communication
abilities.
This
discussion
explores
established
options
like
Augmentative
Alternative
Communication
(AAC)
devices
cutting-edge
approaches
Brain-Computer
Interfaces
(BCIs).
Neuroprosthetics,
field
merging
neuroscience
biomedical
engineering,
aims
replace
or
modulate
damaged
parts
of
the
nervous
system.
We
delved
into
potential
optogenetics,
neuroengineering
devices,
biosensors,
highlighting
contributions
future
advancements
in
speech
assistance.
The
dialogue
emphasized
importance
comprehensive
approach,
combining
with
ongoing
research
on
areas
synthetic
biology.
multi-faceted
approach
holds
key
empowering
speech-impaired
individuals
fostering
improved
communication.
Язык: Английский
Effects of different contents yttrium doping on the electrical performance and stability of bilayer IZO/IYZO thin-film transistors
Applied Physics Letters,
Год журнала:
2025,
Номер
126(14)
Опубликована: Апрель 1, 2025
In
this
study,
the
effects
of
doping
with
different
contents
rare-earth
element
yttrium
(Y)
on
electrical
performance
indium
zinc
oxide
(IZO)
and
bilayer
IZO/IYZO
thin-film
transistors
(TFTs)
were
investigated.
Through
a
rational
design,
TFT
exhibits
best
Vth
0.2
V
μFE
32.6
cm2/V
s.
X-ray
photoelectron
spectroscopy
band
structure
analysis
demonstrated
that,
in
TFT,
6-nm
highly
conductive
ultrathin
IZO
provided
free
electrons
electron
transfer
from
layer
to
IYZO
formation
bending,
thus
increased
carrier
mobility
devices.
addition,
20-nm-thick
controlled
Ne
by
forming
potential
barrier
(0.38
eV),
thereby
increasing
stability.
Moreover,
experimental
characterization
revealed
that
Y
can
reduce
concentration,
oxygen
vacancies,
surface
defects,
other
trap
densities
Therefore,
stability
small
shifts
−0.9,
0.8,
−1.1,
1.0
under
negative
gate
bias
stress,
positive
light
illumination
stress
measurements.
Overall,
designed
TFTs
open
effective
pathways
for
achieving
high-performance
TFTs.
Язык: Английский
Energy‐Efficient DUV Light Programming Nonvolatile Memory Based on MoS2/β‐Ga2O3/MLG Heterostructures for Neuromorphic Computing
Advanced Functional Materials,
Год журнала:
2025,
Номер
unknown
Опубликована: Май 6, 2025
Abstract
As
one
of
the
key
components
for
brain‐inspired
computing,
optoelectrical
synapses
based
on
memory
devices
are
capturing
growing
attention
due
to
their
integrated
function
sense
and
memory.
However,
high
power
consumption
(large
programming
voltage,
optical
density)
difficulty
avoiding
interference
from
solar
radiation
potentially
limiting
applications
in
artificial
neural
systems.
Here,
a
novel
floating
gate
molybdenum
disulfide
(MoS
2
)/β‐phase
Gallium
oxide
(β‐Ga
O
3
)/Multilayer
graphene
is
proposed.
Benefitting
unique
photosensitive
dielectric
properties
β‐Ga
,
device
exhibits
an
excellent
current
switching
ratio
10
6
only
at
low
programming/erasing
voltage
±40
V.
Furthermore,
possesses
strong
anti‐interference
ability
which
can
operate
solar‐blind
region
(254
nm)
with
254.4
pJ
(0
V).
Due
large
single
photon
energy
short
wavelength,
number
(3.25
×
8
)
used
per
operation
much
smaller
than
that
most
present
works.
The
network
model
constructed
achieves
accuracy
91.07%
image
recognition.
These
results
suggest
feasibility
constructing
energy‐efficient,
van
der
Waals
heterostructures
future
visual
neuromorphic
Язык: Английский
Dye Molecule-Induced Optoelectronic Synaptic Behaviors of Monolayer MoSe2
ACS Applied Materials & Interfaces,
Год журнала:
2024,
Номер
17(1), С. 1460 - 1468
Опубликована: Дек. 28, 2024
Although
MoSe2-based
photodetectors
have
achieved
excellent
performance,
the
ultrafast
photoresponse
has
limited
their
application
as
an
optoelectronic
synapse.
In
this
paper,
enhancement
of
rhodamine
6G
molecule
on
memory
time
MoSe2
is
reported.
It
found
that
monolayer
can
be
obviously
enhanced
after
assembly
with
exhibiting
synaptic
characteristics
in
comparison
to
pristine
MoSe2.
Furthermore,
functions,
including
excitatory
postsynaptic
current,
pair-pulse
facilitation,
short-term
memory,
long-term
"learning-experience"
behavior,
and
tunable
learning
forgetting
process,
successfully.
The
distinctive
energy-level
structure
R6G
its
light
absorption
properties
give
access
additional
source
electrons,
thus
enabling
proposed
MoSe2/rhodamine
hybrid
heterostructure
synapse
provide
efficient
protocol
for
realizing
neuromorphic
computing
enormously
facilitate
advancement
electronics.
Язык: Английский