Machine Learning Science and Technology,
Journal Year:
2024,
Volume and Issue:
5(3), P. 035076 - 035076
Published: Sept. 1, 2024
Abstract
Gravitational
lensing
data
is
frequently
collected
at
low
resolution
due
to
instrumental
limitations
and
observing
conditions.
Machine
learning-based
super-resolution
techniques
offer
a
method
enhance
the
of
these
images,
enabling
more
precise
measurements
effects
better
understanding
matter
distribution
in
system.
This
enhancement
can
significantly
improve
our
knowledge
mass
within
galaxy
its
environment,
as
well
properties
background
source
being
lensed.
Traditional
typically
learn
mapping
function
from
lower-resolution
higher-resolution
samples.
However,
methods
are
often
constrained
by
their
dependence
on
optimizing
fixed
distance
function,
which
result
loss
intricate
details
crucial
for
astrophysical
analysis.
In
this
work,
we
introduce
DiffLense
,
novel
pipeline
based
conditional
diffusion
model
specifically
designed
gravitational
images
obtained
Hyper
Suprime-Cam
Subaru
Strategic
Program
(HSC-SSP).
Our
approach
adopts
generative
model,
leveraging
detailed
structural
information
present
Hubble
space
telescope
(HST)
counterparts.
The
trained
generate
HST
data,
conditioned
HSC
pre-processed
with
denoising
thresholding
reduce
noise
interference.
process
leads
distinct
less
overlapping
during
model’s
training
phase.
We
demonstrate
that
outperforms
existing
state-of-the-art
single-image
techniques,
particularly
retaining
fine
necessary
analyses.
Journal of Physics D Applied Physics,
Journal Year:
2024,
Volume and Issue:
57(49), P. 493004 - 493004
Published: Aug. 14, 2024
Abstract
Chiroptical
metamaterials
have
attracted
considerable
attention
owing
to
their
exciting
opportunities
for
fundamental
research
and
practical
applications
over
the
past
20
years.
Through
designs,
chiroptical
response
of
chiral
can
be
several
orders
magnitude
higher
than
that
natural
materials.
therefore
represent
a
special
type
artificial
structures
unique
activities.
In
this
review,
we
present
comprehensive
overview
progresses
in
development
metamaterials.
metamaterial
progress
enables
applications,
including
asymmetric
transmission,
polarization
conversion,
absorber,
imaging,
sensor
emission.
We
also
review
fabrication
techniques
design
based
on
deep
learning.
conclusion,
possible
further
directions
field.
Opto-Electronic Advances,
Journal Year:
2023,
Volume and Issue:
6(10), P. 230057 - 230057
Published: Jan. 1, 2023
Chiral
nanostructures
can
enhance
the
weak
inherent
chiral
effects
of
biomolecules
and
highlight
important
roles
in
detection.
However,
design
is
challenged
by
extensive
theoretical
simulations
explorative
experiments.
Recently,
Zheyu
Fang's
group
proposed
a
nanostructure
method
based
on
reinforcement
learning,
which
find
out
metallic
with
sharp
peak
circular
dichroism
spectra
detection
signals.
This
work
envisions
powerful
artificial
intelligence
nanophotonic
designs.
IEEE Microwave and Wireless Technology Letters,
Journal Year:
2024,
Volume and Issue:
34(5), P. 467 - 470
Published: April 2, 2024
The
neuro-coupled
mode
theory
(i.e.,
neuro-CMT)
approach
has
been
recently
reported
for
the
intelligent
design
of
metasurfaces.
This
letter
presents
an
advance,
that
is,
feature-assisted
neuro-CMT
approach,
to
address
issue
bad
starting
points
and
increase
optimization
efficiency
further.
We
define
resonant
frequencies
in
original
surrogate
as
feature
parameters
identify
them
additional
outputs.
Then,
we
formulate
a
feature-based
objective
function
guide
automatically
move
into
desired
frequency
band
at
initial
stage,
while
ensuring
electromagnetic
(EM)
response
meets
specification
subsequent
process.
proposed
is
applied
two
metasurface
microwave
absorbers,
showing
increased
convergence
speed
solution
optimality
compared
with
existing
approach.
Numerical
simulations
experimental
measurements
further
verify
accuracy
Machine Learning Science and Technology,
Journal Year:
2024,
Volume and Issue:
5(3), P. 035076 - 035076
Published: Sept. 1, 2024
Abstract
Gravitational
lensing
data
is
frequently
collected
at
low
resolution
due
to
instrumental
limitations
and
observing
conditions.
Machine
learning-based
super-resolution
techniques
offer
a
method
enhance
the
of
these
images,
enabling
more
precise
measurements
effects
better
understanding
matter
distribution
in
system.
This
enhancement
can
significantly
improve
our
knowledge
mass
within
galaxy
its
environment,
as
well
properties
background
source
being
lensed.
Traditional
typically
learn
mapping
function
from
lower-resolution
higher-resolution
samples.
However,
methods
are
often
constrained
by
their
dependence
on
optimizing
fixed
distance
function,
which
result
loss
intricate
details
crucial
for
astrophysical
analysis.
In
this
work,
we
introduce
DiffLense
,
novel
pipeline
based
conditional
diffusion
model
specifically
designed
gravitational
images
obtained
Hyper
Suprime-Cam
Subaru
Strategic
Program
(HSC-SSP).
Our
approach
adopts
generative
model,
leveraging
detailed
structural
information
present
Hubble
space
telescope
(HST)
counterparts.
The
trained
generate
HST
data,
conditioned
HSC
pre-processed
with
denoising
thresholding
reduce
noise
interference.
process
leads
distinct
less
overlapping
during
model’s
training
phase.
We
demonstrate
that
outperforms
existing
state-of-the-art
single-image
techniques,
particularly
retaining
fine
necessary
analyses.
