Mathematical and Computational Applications,
Journal Year:
2024,
Volume and Issue:
29(4), P. 56 - 56
Published: July 13, 2024
Feature
selection
is
a
preprocessing
step
in
machine
learning
that
aims
to
reduce
dimensionality
and
improve
performance.
The
approaches
for
feature
are
often
classified
according
the
evaluation
of
subset
features
as
filter,
wrapper,
embedded
approaches.
high
performance
wrapper
associated
at
same
time
with
disadvantage
computational
cost.
Cost-reduction
mechanisms
have
been
proposed
literature,
where
competitive
achieved
more
efficiently.
This
work
applies
simple
effective
resource-saving
fixed
incremental
sampling
fraction
strategies
memory
avoid
repeated
evaluations
multi-objective
permutational-based
differential
evolution
selection.
selected
approach
an
extension
DE-FSPM
algorithm
mechanism
GDE3
algorithm.
results
showed
resource
savings,
especially
number
required
search
process.
Nonetheless,
it
was
also
detected
algorithm’s
diminished.
Therefore,
reported
literature
on
effectiveness
cost
reduction
single-objective
were
only
partially
sustained
Sensors,
Journal Year:
2025,
Volume and Issue:
25(1), P. 228 - 228
Published: Jan. 3, 2025
Recent
advancements
in
Earth
Observation
sensors,
improved
accessibility
to
imagery
and
the
development
of
corresponding
processing
tools
have
significantly
empowered
researchers
extract
insights
from
Multisource
Remote
Sensing.
This
study
aims
use
these
technologies
for
mapping
summer
winter
Land
Use/Land
Cover
features
Cuenca
de
la
Laguna
Merín,
Uruguay,
while
comparing
performance
Random
Forests,
Support
Vector
Machines,
Gradient-Boosting
Tree
classifiers.
The
materials
include
Sentinel-2,
Sentinel-1
Shuttle
Radar
Topography
Mission
imagery,
Google
Engine,
training
validation
datasets
quoted
methods
involve
creating
a
multisource
database,
conducting
feature
importance
analysis,
developing
models,
supervised
classification
performing
accuracy
assessments.
Results
indicate
low
significance
microwave
inputs
relative
optical
features.
Short-wave
infrared
bands
transformations
such
as
Normalised
Vegetation
Index,
Surface
Water
Index
Enhanced
demonstrate
highest
importance.
Accuracy
assessments
that
various
classes
is
optimal,
particularly
rice
paddies,
which
play
vital
role
country’s
economy
highlight
significant
environmental
concerns.
However,
challenges
persist
reducing
confusion
between
classes,
regarding
natural
vegetation
versus
seasonally
flooded
vegetation,
well
post-agricultural
fields/bare
land
herbaceous
areas.
Forests
Trees
exhibited
superior
compared
Machines.
Future
research
should
explore
approaches
Deep
Learning
pixel-based
object-based
integration
address
identified
challenges.
These
initiatives
consider
data
combinations,
including
additional
indices
texture
metrics
derived
Grey-Level
Co-Occurrence
Matrix.
Mathematics,
Journal Year:
2025,
Volume and Issue:
13(5), P. 833 - 833
Published: March 2, 2025
In
response
to
the
limitations
of
reinforcement
learning
and
Evolutionary
Algorithms
(EAs)
in
complex
problem-solving,
Reinforcement
Learning
(EvoRL)
has
emerged
as
a
synergistic
solution.
This
systematic
review
aims
provide
comprehensive
analysis
EvoRL,
examining
symbiotic
relationship
between
EAs
algorithms
identifying
critical
gaps
relevant
application
tasks.
The
begins
by
outlining
technological
foundations
detailing
complementary
address
learning,
such
parameter
sensitivity,
sparse
rewards,
its
susceptibility
local
optima.
We
then
delve
into
challenges
faced
both
exploring
utility
EvoRL.
EvoRL
itself
is
constrained
sampling
efficiency
algorithmic
complexity,
which
affect
areas
like
robotic
control
large-scale
industrial
settings.
Furthermore,
we
significant
open
issues
field,
adversarial
robustness,
fairness,
ethical
considerations.
Finally,
propose
future
directions
for
emphasizing
research
avenues
that
strive
enhance
self-adaptation,
self-improvement,
scalability,
interpretability,
so
on.
To
quantify
current
state,
analyzed
about
100
studies,
categorizing
them
based
on
algorithms,
performance
metrics,
benchmark
Serving
resource
researchers
practitioners,
this
provides
insights
state
offers
guide
advancing
capabilities
ever-evolving
landscape
artificial
intelligence.
Automated Software Engineering,
Journal Year:
2025,
Volume and Issue:
32(2)
Published: March 15, 2025
Abstract
Automated
program
repair
techniques
aim
to
aid
software
developers
with
the
challenging
task
of
fixing
bugs.
In
heuristic-based
repair,
a
search
space
mutated
variants
is
explored
find
potential
patches
for
Most
commonly,
every
selection
mutation
operator
during
performed
uniformly
at
random,
which
can
generate
many
buggy,
even
uncompilable
programs.
Our
goal
reduce
generation
that
do
not
compile
or
break
intended
functionality
waste
considerable
resources.
this
paper,
we
investigate
feasibility
reinforcement
learning-based
approach
operators
in
repair.
proposed
programming
language,
granularity-level,
and
strategy
agnostic
allows
easy
augmentation
into
existing
tools.
We
conducted
an
extensive
empirical
evaluation
four
techniques,
two
reward
types,
credit
assignment
strategies,
integration
methods,
three
sets
using
30,080
independent
attempts.
evaluated
our
on
353
real-world
bugs
from
Defects4J
benchmark.
The
results
higher
number
test-passing
variants,
but
does
exhibit
noticeable
improvement
patched
comparison
baseline,
uniform
random
selection.
While
learning
has
been
previously
shown
be
successful
improving
evolutionary
algorithms,
often
used
it
yet
demonstrate
such
improvements
when
applied
area
research.
Journal of Marine Science and Engineering,
Journal Year:
2025,
Volume and Issue:
13(3), P. 596 - 596
Published: March 17, 2025
In
ship
navigation,
determining
a
safe
and
economic
path
from
start
to
destination
under
dynamic
complex
environment
is
essential,
but
the
traditional
algorithms
of
current
research
are
inefficient.
Therefore,
novel
differential
evolution
deep
reinforcement
learning
algorithm
(DEDRL)
proposed
address
problems,
which
composed
local
planning
global
planning.
The
Deep
Q-Network
utilized
search
best
in
target
multiple-obstacles
scenarios.
Furthermore,
course-punishing
reward
mechanism
introduced
optimize
constrain
detected
length
as
short
possible.
Quaternion
domain
COLREGs
involved
construct
collision
risk
detection
model.
Compared
with
other
algorithms,
experimental
results
demonstrate
that
DEDRL
achieved
28.4539
n
miles,
also
performed
all
scenarios
Overall,
reliable
robust
for
it
provides
an
efficient
solution
avoidance.
