Biomimetics,
Год журнала:
2025,
Номер
10(6), С. 361 - 361
Опубликована: Июнь 3, 2025
Hyper-parameters
play
a
critical
role
in
neural
networks;
they
significantly
impact
both
training
effectiveness
and
overall
model
performance.
Proper
hyper-parameter
settings
can
accelerate
convergence
improve
generalization.
Among
various
hyper-parameters,
the
learning
rate
is
particularly
important.
However,
optimizing
typically
requires
extensive
experimentation
tuning,
as
its
setting
often
dependent
on
specific
tasks
datasets
therefore
lacks
universal
rules
or
standards.
Consequently,
adjustments
are
generally
made
through
trial
error,
thereby
making
selection
of
complex
time-consuming.
In
an
attempt
to
surmount
this
challenge,
evolutionary
computation
algorithms
automatically
adjust
efficiency
response
this,
we
propose
black
widow
optimization
algorithm
based
Lagrange
interpolation
(LIBWONN)
optimize
ResNet18.
Moreover,
evaluate
LIBWONN’s
using
24
benchmark
functions
from
CEC2017
CEC2022
compare
it
with
nine
advanced
metaheuristic
algorithms.
The
experimental
results
indicate
that
LIBWONN
outperforms
other
stability.
Additionally,
experiments
publicly
available
six
different
fields
demonstrate
improves
accuracy
testing
sets
compared
standard
BWO,
gains
6.99%
4.48%,
respectively.
Scientific Reports,
Год журнала:
2025,
Номер
15(1)
Опубликована: Апрель 10, 2025
Temperature
regulation
in
nonlinear
and
highly
dynamic
processes
such
as
the
continuous
stirred-tank
heater
(CSTH)
is
a
challenging
task
due
to
inherent
system
nonlinearities
disturbances.
This
study
proposes
novel
metaheuristic-driven
control
strategy,
combining
two
degrees
of
freedom-PID
acceleration
(2DOF-PIDA)
controller
with
recently
developed
starfish
optimization
algorithm
(SFOA)
for
temperature
CSTH
process.
The
2DOF-PIDA
enhances
performance
by
decoupling
setpoint
tracking
disturbance
rejection,
while
SFOA
ensures
optimal
tuning
parameters
leveraging
its
powerful
exploration
exploitation
capabilities.
Simulation
results
validate
effectiveness
proposed
approach,
demonstrating
improved
accuracy,
robustness
compared
conventional
methods.
combination
provides
flexible
efficient
solution
controlling
systems,
significant
implications
industrial
applications.
Scientific Reports,
Год журнала:
2025,
Номер
15(1)
Опубликована: Фев. 24, 2025
Practical
engineering
optimization
problems
are
characterized
by
high
dimensionality,
non-convexity,
and
non-linearity,
the
use
of
optimizers
to
provide
better
quality
solutions
target
problem
in
an
acceptable
time
is
a
hot
research
topic
field
optimal
design.
In
this
paper,
inspired
Sturnus
vulgaris
escape
behavior,
Vulgaris
Escape
Algorithm
(SVEA)
proposed
high-performance
optimizer
for
complex
problems.
The
algorithm
composed
exploration
exploitation
strategies,
controlled
fixed
parameters.
strategies
include
High-Altitude
Strategy
Wave
1,
while
consist
Cordon
Line
2.
enhances
capabilities
reorganizing
subgroups,
preventing
leader
individuals
from
overlapping,
avoiding
collisions
between
individuals.
conducts
refined
searches
around
high-value
regions,
further
improving
precision.
Strategies
1
2
help
population
local
optima
prevent
over-spreading.
performance
SVEA
evaluated
through
employment
23
benchmark
test
functions
CEC2017
set,
with
subsequent
comparison
undertaken
nine
statE
−
of-thE
art
meta-heuristic
algorithms.
outcomes
evaluation
demonstrate
that
attains
top
ranking
identified
as
best-performing
across
all
sets.
A
statistical
analysis
reveals
solution
set
exhibits
superior
other
algorithms,
discrepancy
being
deemed
be
statistically
significant.
Finally,
applied
five
real-world
problems,
providing
satisfying
constraints.
Journal of Marine Science and Engineering,
Год журнала:
2025,
Номер
13(4), С. 787 - 787
Опубликована: Апрель 15, 2025
Sound
speed
profiles
(SSPs)
must
be
detected
simultaneously
to
perform
a
multibeam
depth
survey.
Accurate
real-time
sound
profile
(SSP)
acquisition
remains
critical
challenge
in
deep-sea
bathymetry
due
the
limitations
regarding
direct
measurements
under
harsh
operational
conditions.
To
address
issue,
we
propose
joint
inversion
framework
integrating
World
Ocean
Atlas
2023
(WOA23)
temperature–salinity
model
data,
historical
situ
SSPs,
and
surface
measurements.
By
constructing
high-resolution
regional
field
through
WOA23
SSP
fusion,
this
method
effectively
mitigates
spatiotemporal
heterogeneity
seasonal
variability.
The
artificial
lemming
algorithm
(ALA)
is
introduced
optimize
of
empirical
orthogonal
function
(EOF)
coefficients,
enhancing
global
search
efficiency
while
avoiding
local
optimization.
An
experimental
validation
northwest
Pacific
demonstrated
that
proposed
has
better
performance
than
conventional
substitution,
interpolation,
WOA23-only
approaches.
results
indicate
mean
absolute
error
(MAE),
root
square
(RMSE),
maximum
(ME)
reconstruction
are
reduced
by
41.5%,
46.0%,
49.4%,
respectively.
When
reconstructed
SSPs
applied
bathymetric
correction,
errors
further
0.193
m
0.213
0.394
(ME),
suppressing
“smiley
face”
distortion
caused
gradient
anomalies.
dynamic
selection
first
six
EOF
modes
balances
computational
fidelity.
This
study
provides
robust
solution
for
estimation
data-scarce
environments,
particularly
underwater
autonomous
vehicles.
seabed
missing
significantly
accuracy
surveys.
Biomimetics,
Год журнала:
2025,
Номер
10(5), С. 260 - 260
Опубликована: Апрель 23, 2025
In
real-world
applications,
many
complex
problems
can
be
formulated
as
mathematical
optimization
challenges,
and
efficiently
solving
these
is
critical.
Metaheuristic
algorithms
have
proven
highly
effective
in
addressing
a
wide
range
of
engineering
issues.
The
differentiated
creative
search
recently
proposed
evolution-based
meta-heuristic
algorithm
with
certain
advantages.
However,
it
also
has
limitations,
including
weakened
population
diversity,
reduced
efficiency,
hindrance
comprehensive
exploration
the
solution
space.
To
address
shortcomings
DCS
algorithm,
this
paper
proposes
multi-strategy
(MSDCS)
based
on
collaborative
development
mechanism
evaluation
strategy.
First,
that
organically
integrates
estimation
distribution
to
compensate
for
algorithm's
insufficient
ability
its
tendency
fall
into
local
optimums
through
guiding
effect
dominant
populations,
improve
quality
efficiency
at
same
time.
Secondly,
new
strategy
realize
coordinated
transition
between
exploitation
fitness
distance.
Finally,
linear
size
reduction
incorporated
DCS,
which
significantly
improves
overall
performance
by
maintaining
large
initial
stage
enhance
capability
extensive
space,
then
gradually
decreasing
later
capability.
A
series
validations
was
conducted
CEC2018
test
set,
experimental
results
were
analyzed
using
Friedman
Wilcoxon
rank
sum
test.
show
superior
MSDCS
terms
convergence
speed,
stability,
global
optimization.
addition,
successfully
applied
several
constrained
problems.
all
cases,
outperforms
basic
fast
strong
robustness,
emphasizing
efficacy
practical
applications.
Agronomy,
Год журнала:
2025,
Номер
15(5), С. 1057 - 1057
Опубликована: Апрель 27, 2025
The
quality
of
medicinal
plants
is
closely
related
to
the
ecological
factors
their
growing
environment,
as
efficacy
reflected
in
content
key
components,
which
turn
indicates
plants.
This
study
measured
daily
variations
major
constituents,
including
lobetyolin,
polysaccharides,
and
total
flavonoids,
Codonopsis
pilosula
(Franch.)
Nannf.,
Changzhi
Jincheng
regions
Shanxi
Province,
China
known
Lu
Tangshen.
Throughout
its
growth
cycle.
Additionally,
explored
effects
11
(both
climatic
soil
variables)
on
primary
components
C.
pilosula.
Through
block
experiments
comparisons
between
future
data
predictions
actual
measurements,
reliability
model
consistency
experimental
were
ultimately
confirmed.
Principal
component
analysis
(PCA),
stepwise
multiple
linear
regression
analysis,
nonlinear
polynomial
modeling
employed
investigate
relationships
quality-related
constituents
(polysaccharides,
lobetyolin).
results
showed
that
models
effectively
explained
temperature
(DT)
with
an
adjusted
R2
exceeding
0.8,
but
due
inherently
nature
data,
it
evident
are
fundamentally
inadequate
for
accurately
capturing
underlying
relationships.
Therefore,
fit
flavonoids
lobetyolin
was
suboptimal.
introduction
(second-,
fourth-,
fifth-order)
significantly
improved
fit,
indicating
existence
complex
components.
For
fourth-order
demonstrated
best
performance,
while
fifth-order
required
adequately
describe
lobetyolin.
Based
models,
optimal
ranges
identified:
polysaccharides
influenced
by
atmospheric
pressure
(AP)
9.1
9.3
kPa,
air
relative
humidity
(ARH)
30%
60%,
40
cm
mean
annual
(40cmMAT)
27.5
°C
28.5
°C,
pH
9.68
9.72,
nitrogen
(N)
7
9
mg/kg.
narrow
observed
temperature,
humidity,
(MAT
10
15
40cmMAT
9.72).
Lobetyolin
conditions
at
AP
28.0
ARH
65%
75%,
near
9.70,
days
after
planting
(DAP)
50.
adoption
higher-order
clarified
critical
inflection
points
ranges,
providing
a
refined
reference
enhancing
These
findings
offer
valuable
insights
precision
cultivation
strategies
aimed
improving
Scientific Reports,
Год журнала:
2025,
Номер
15(1)
Опубликована: Май 2, 2025
Magnetic
target
state
estimation
is
a
widely
applied
technology,
but
it
also
faces
many
challenges
in
practical
applications.
One
of
the
most
critical
issue
accuracy.
The
Grey
Wolf
Optimizer
(GWO)
one
more
successful
swarm
intelligence
algorithms
recent
years,
its
shortcomings
have
been
exposed
when
facing
increasingly
complex
problems.
Therefore,
Multi-Strategy
Improved
(MSIGWO)
algorithm
has
proposed
to
enhance
accuracy
magnetic
estimation.
In
initialization
phase,
Tent
chaos
mapping
introduced
population
diversity,
prevent
falling
into
local
optima,
and
improve
convergence
speed.
Multi-population
fusion
evolution
strategies
accuracy,
global
search
ability.
Nonlinear
factors
better
balance
exploration
exploitation
behaviors.
Dynamic
weight
increase
diversity
samples
reduce
likelihood
optima.
Adaptive
dimensional
learning
balances
searches,
enhancing
diversity.
Levy
flight
enhances
ability
jump
out
optima
ensures
CEC2018
benchmark
function
set
29
problems
problems,
MSIGWO
was
tested,
statistical
indicators
Friedman
test
results
show
that
compared
with
GWO
advanced
variants,
superior
performance.
application
this
proven
effectiveness
applicability.
Aerospace,
Год журнала:
2025,
Номер
12(5), С. 413 - 413
Опубликована: Май 7, 2025
The
vigorous
development
of
urban
air
mobility
(UAM)
is
reshaping
the
travel
landscape,
but
it
also
poses
severe
challenges
to
safe
and
efficient
operation
dense
complex
airspace.
Potential
conflicts
between
flight
plans
have
become
a
core
bottleneck
restricting
its
development.
Traditional
plan
adjustment
management
methods
often
rely
on
deterministic
trajectory
predictions,
ignoring
inherent
temporal
uncertainties
in
actual
operations,
which
may
lead
underestimation
potential
risks.
Meanwhile,
existing
global
optimization
strategies
face
issues
inefficiency
overly
broad
scopes
when
dealing
with
large-scale
conflicts.
To
address
these
challenges,
this
study
proposes
an
innovative
conflict
framework.
First,
by
introducing
probabilistic
model
time
errors,
new
detection
mechanism
based
confidence
intervals
constructed,
significantly
enhancing
ability
foresee
non-obvious
Furthermore,
network
theory,
framework
accurately
identifies
small
number
“critical
plans”
that
play
role
network,
revealing
their
key
impact
chain
reactions
On
basis,
phased
strategy
adopted,
prioritizing
spatiotemporal
parameters
(departure
speed)
for
critical
systematically
resolve
most
Subsequently,
only
fine-tuning
speeds
non-critical
required
remaining
local
conflicts,
thereby
minimizing
interference
overall
operational
order.
Simulation
results
demonstrate
not
improves
comprehensiveness
effectively
reduces
total
Additionally,
proposed
artificial
lemming
algorithm
(ALA)
outperforms
traditional
algorithms
terms
solution
quality.
This
work
provides
important
theoretical
foundation
practically
valuable
developing
robust
UAM
dynamic
scheduling
systems,
holding
promise
support
orderly
traffic
future.