Mathematics,
Год журнала:
2022,
Номер
10(16), С. 3019 - 3019
Опубликована: Авг. 22, 2022
Deep
learning
has
been
widely
used
in
different
fields
such
as
computer
vision
and
speech
processing.
The
performance
of
deep
algorithms
is
greatly
affected
by
their
hyperparameters.
For
complex
machine
models
neural
networks,
it
difficult
to
determine
In
addition,
existing
hyperparameter
optimization
easily
converge
a
local
optimal
solution.
This
paper
proposes
method
for
that
combines
the
Sparrow
Search
Algorithm
Particle
Swarm
Optimization,
called
Hybrid
Algorithm.
takes
advantages
avoiding
solution
search
efficiency
Optimization
achieve
global
optimization.
Experiments
verified
proposed
algorithm
simple
networks.
results
show
strong
capability
avoid
solutions
satisfactory
both
low
high-dimensional
spaces.
provides
new
problems
models.
Journal Of Big Data,
Год журнала:
2023,
Номер
10(1)
Опубликована: Апрель 14, 2023
Abstract
Data
scarcity
is
a
major
challenge
when
training
deep
learning
(DL)
models.
DL
demands
large
amount
of
data
to
achieve
exceptional
performance.
Unfortunately,
many
applications
have
small
or
inadequate
train
frameworks.
Usually,
manual
labeling
needed
provide
labeled
data,
which
typically
involves
human
annotators
with
vast
background
knowledge.
This
annotation
process
costly,
time-consuming,
and
error-prone.
every
framework
fed
by
significant
automatically
learn
representations.
Ultimately,
larger
would
generate
better
model
its
performance
also
application
dependent.
issue
the
main
barrier
for
dismissing
use
DL.
Having
sufficient
first
step
toward
any
successful
trustworthy
application.
paper
presents
holistic
survey
on
state-of-the-art
techniques
deal
models
overcome
three
challenges
including
small,
imbalanced
datasets,
lack
generalization.
starts
listing
techniques.
Next,
types
architectures
are
introduced.
After
that,
solutions
address
listed,
such
as
Transfer
Learning
(TL),
Self-Supervised
(SSL),
Generative
Adversarial
Networks
(GANs),
Model
Architecture
(MA),
Physics-Informed
Neural
Network
(PINN),
Deep
Synthetic
Minority
Oversampling
Technique
(DeepSMOTE).
Then,
these
were
followed
some
related
tips
about
acquisition
prior
purposes,
well
recommendations
ensuring
trustworthiness
dataset.
The
ends
list
that
suffer
from
scarcity,
several
alternatives
proposed
in
order
more
each
Electromagnetic
Imaging
(EMI),
Civil
Structural
Health
Monitoring,
Medical
imaging,
Meteorology,
Wireless
Communications,
Fluid
Mechanics,
Microelectromechanical
system,
Cybersecurity.
To
best
authors’
knowledge,
this
review
offers
comprehensive
overview
strategies
tackle
Electronics,
Год журнала:
2022,
Номер
11(22), С. 3798 - 3798
Опубликована: Ноя. 18, 2022
Developing
countries
have
had
numerous
obstacles
in
diagnosing
the
COVID-19
worldwide
pandemic
since
its
emergence.
One
of
most
important
ways
to
control
spread
this
disease
begins
with
early
detection,
which
allows
that
isolation
and
treatment
could
perhaps
be
started.
According
recent
results,
chest
X-ray
scans
provide
information
about
onset
infection,
may
evaluated
so
diagnosis
can
begin
sooner.
This
is
where
artificial
intelligence
collides
skilled
clinicians’
diagnostic
abilities.
The
suggested
study’s
goal
make
a
contribution
battling
epidemic
by
using
simple
convolutional
neural
network
(CNN)
model
construct
an
automated
image
analysis
framework
for
recognizing
afflicted
data.
To
improve
classification
accuracy,
fully
connected
layers
CNN
were
replaced
efficient
extreme
gradient
boosting
(XGBoost)
classifier,
used
categorize
extracted
features
layers.
Additionally,
hybrid
version
arithmetic
optimization
algorithm
(AOA),
also
developed
facilitate
proposed
research,
tune
XGBoost
hyperparameters
images.
Reported
experimental
data
showed
approach
outperforms
other
state-of-the-art
methods,
including
cutting-edge
metaheuristics
algorithms,
tested
same
framework.
For
validation
purposes,
balanced
images
dataset
12,000
observations,
belonging
normal,
viral
pneumonia
classes,
was
used.
method,
tuned
introduced
AOA,
superior
performance,
achieving
accuracy
approximately
99.39%
weighted
average
precision,
recall
F1-score
0.993889,
0.993887
0.993887,
respectively.
Heliyon,
Год журнала:
2024,
Номер
10(5), С. e26799 - e26799
Опубликована: Фев. 28, 2024
Computer-aided
diagnosis
(CAD)
systems
play
a
vital
role
in
modern
research
by
effectively
minimizing
both
time
and
costs.
These
support
healthcare
professionals
like
radiologists
their
decision-making
process
efficiently
detecting
abnormalities
as
well
offering
accurate
dependable
information.
heavily
depend
on
the
efficient
selection
of
features
to
accurately
categorize
high-dimensional
biological
data.
can
subsequently
assist
related
medical
conditions.
The
task
identifying
patterns
biomedical
data
be
quite
challenging
due
presence
numerous
irrelevant
or
redundant
features.
Therefore,
it
is
crucial
propose
then
utilize
feature
(FS)
order
eliminate
these
primary
goal
FS
approaches
improve
accuracy
classification
eliminating
that
are
less
informative.
phase
plays
critical
attaining
optimal
results
machine
learning
(ML)-driven
CAD
systems.
effectiveness
ML
models
significantly
enhanced
incorporating
during
training
phase.
This
empirical
study
presents
methodology
for
using
technique.
proposed
approach
incorporates
three
soft
computing-based
optimization
algorithms,
namely
Teaching
Learning-Based
Optimization
(TLBO),
Elephant
Herding
(EHO),
hybrid
algorithm
two.
algorithms
were
previously
employed;
however,
addressing
issues
predicting
human
diseases
has
not
been
investigated.
following
evaluation
focuses
categorization
benign
malignant
tumours
publicly
available
Wisconsin
Diagnostic
Breast
Cancer
(WDBC)
benchmark
dataset.
five-fold
cross-validation
technique
employed
mitigate
risk
over-fitting.
approach's
proficiency
determined
based
several
metrics,
including
sensitivity,
specificity,
precision,
accuracy,
area
under
receiver-operating
characteristic
curve
(AUC),
F1-score.
best
value
computed
through
suggested
97.96%.
clinical
decision
system
demonstrates
highly
favourable
performance
outcome,
making
valuable
tool
practitioners
secondary
opinion
reducing
overburden
expert
practitioners.
IEEE Access,
Год журнала:
2022,
Номер
10, С. 10907 - 10933
Опубликована: Янв. 1, 2022
Many
new
algorithms
have
been
proposed
to
solve
the
mathematical
equations
formulated
describe
real-world
problems.
But
there
still
does
not
exist
one
algorithm
that
could
problems
all.
And
most
of
defects
in
some
aspects,
they
need
be
improved
application.
In
order
find
a
more
efficient
optimization
and
inspired
by
better
performance
Arithmetic
Optimization
(AOA)
Aquila
Optimizer
(AO),
we
hybridization
them
abbreviated
AOAAO
this
paper.
Considering
Harris
Hawk
(HHO)
algorithm,
an
energy
parameter
E
was
also
introduced
balance
exploration
exploitation
procedures
individuals
swarms,
furthermore,
piecewise
linear
map
decrease
randomness
parameter.
Pseudo
code
presented,
Simulation
experiments
were
carried
out
on
benchmark
functions
three
classical
engineering
involved
optimization.
Nine
popular
well
demonstrated
included
for
comparison.
Results
confirmed
would
with
faster
convergence
rate,
higher
accuracy.