Physical review. E,
Год журнала:
2023,
Номер
108(6)
Опубликована: Дек. 21, 2023
The
reservoir
computing
approach
utilizes
a
time
series
of
measurements
as
input
to
high-dimensional
dynamical
system
known
reservoir.
However,
the
relies
on
sampling
random
matrix
define
its
underlying
layer,
which
leads
numerous
hyperparameters
that
need
be
optimized.
Here,
we
propose
nonlocally
coupled
pendulum
model
with
higher-order
interactions
novel
reservoir,
requires
no
matrices
and
fewer
hyperparameters.
We
use
Bayesian
optimization
explore
hyperparameter
space
within
minimal
number
iterations
train
pendulums
reproduce
chaotic
attractors,
simplifies
complicated
optimization.
illustrate
effectiveness
our
technique
Lorenz
Hindmarsh-Rose
neuronal
model,
calculate
Pearson
correlation
coefficients
between
Hausdorff
metrics
in
phase
space.
demonstrate
contribution
by
analyzing
interaction
different
configurations
prediction
performance,
well
computations
largest
Lyapunov
exponents.
chimera
state
is
found
most
effective
regime
for
prediction.
findings,
where
present
new
structure,
offer
potential
applications
design
high-performance
modeling
dynamics
physical
systems.
Physical review. E,
Год журнала:
2024,
Номер
110(3)
Опубликована: Сен. 10, 2024
We
report
higher-order
coupling
induced
stable
chimeralike
state
in
a
bipartite
network
of
coupled
phase
oscillators
without
any
time-delay
the
coupling.
show
that
interaction
breaks
symmetry
homogeneous
synchronized
to
facilitate
manifestation
breaking
state.
In
particular,
such
manifests
only
when
pairwise
is
attractive
and
repulsive,
vice
versa.
Further,
we
also
demonstrate
increased
degree
heterogeneity
promotes
symmetric
states
diagram
by
suppressing
asymmetric
deduce
low-dimensional
evolution
equations
for
macroscopic
order
parameters
using
Ott-Antonsen
ansatz
obtain
bifurcation
curves
from
them
software
xppaut,
which
agrees
very
well
with
simulation
results.
analytical
stability
conditions
incoherent
state,
in-phase
out-of-phase
states,
match
curves.
Chaos An Interdisciplinary Journal of Nonlinear Science,
Год журнала:
2023,
Номер
33(11)
Опубликована: Ноя. 1, 2023
Higher-order
interactions
improve
our
capability
to
model
real-world
complex
systems
ranging
from
physics
and
neuroscience
economics
social
sciences.
There
is
great
interest
nowadays
in
understanding
the
contribution
of
higher-order
terms
collective
behavior
network.
In
this
work,
we
investigate
stability
complete
synchronization
networks
with
structures.
We
demonstrate
that
level
a
network
composed
nodes
interacting
simultaneously
via
multiple
orders
maintained
regardless
intensity
coupling
strength
across
different
orders.
articulate
lower-order
topologies
work
together
complementarily
provide
optimal
stable
configuration,
challenging
previous
conclusions
promote
synchronization.
Furthermore,
find
simply
adding
based
on
existing
connections,
as
simple
complexes,
does
not
have
significant
impact
The
universal
applicability
lies
comprehensive
analysis
topologies,
including
hypergraphs
simplicial
utilization
appropriate
rescaling
assess
stability.
Physical review. E,
Год журнала:
2023,
Номер
108(6)
Опубликована: Дек. 21, 2023
The
reservoir
computing
approach
utilizes
a
time
series
of
measurements
as
input
to
high-dimensional
dynamical
system
known
reservoir.
However,
the
relies
on
sampling
random
matrix
define
its
underlying
layer,
which
leads
numerous
hyperparameters
that
need
be
optimized.
Here,
we
propose
nonlocally
coupled
pendulum
model
with
higher-order
interactions
novel
reservoir,
requires
no
matrices
and
fewer
hyperparameters.
We
use
Bayesian
optimization
explore
hyperparameter
space
within
minimal
number
iterations
train
pendulums
reproduce
chaotic
attractors,
simplifies
complicated
optimization.
illustrate
effectiveness
our
technique
Lorenz
Hindmarsh-Rose
neuronal
model,
calculate
Pearson
correlation
coefficients
between
Hausdorff
metrics
in
phase
space.
demonstrate
contribution
by
analyzing
interaction
different
configurations
prediction
performance,
well
computations
largest
Lyapunov
exponents.
chimera
state
is
found
most
effective
regime
for
prediction.
findings,
where
present
new
structure,
offer
potential
applications
design
high-performance
modeling
dynamics
physical
systems.
