Deep
Learning-Based
Recommender
Systems
(DLRS)
represent
a
prominent
research
area
in
the
academic
community.
This
paper
aims
to
conduct
bibliometric
and
visualization
analysis
using
VOSviewer
software,
based
on
1,435
DLRS-related
publications
retrieved
from
Web
of
Science
database.
By
analyzing
existing
literature,
this
study
investigates
quantity
DLRS
papers,
their
origins,
affiliated
institutions,
notable
authors.
The
findings
reveal
substantial
rapid
growth
trend
since
2016.
Co-occurrence
uncovers
five
major
directions
within
field:
deep
recommender
systems,
collaborative
prediction
models,
neural
network
personalization,
attention-based
models
recommendation,
learning
for
systems.
Finally,
provides
relevant
recommendations
future
practical
applications
DLRS,
addressing
both
researchers
industry
professionals.
Scientific Reports,
Journal Year:
2024,
Volume and Issue:
14(1)
Published: May 6, 2024
Abstract
In
recent
years,
the
proliferation
of
Massive
Open
Online
Courses
(MOOC)
platforms
on
a
global
scale
has
been
remarkable.
Learners
can
now
meet
their
learning
demands
with
help
MOOC.
However,
learners
might
not
understand
course
material
well
if
they
have
access
to
lot
information
due
inadequate
expertise
and
cognitive
ability.
Personalized
Recommender
Systems
(RSs),
cutting-edge
technology,
assist
in
addressing
this
issue.
It
greatly
increases
resource
acquisition
through
personalized
availability
for
various
people
all
ages.
Intelligent
methods,
such
as
machine
Reinforcement
Learning
(RL)
be
used
RS
challenges.
needs
supervised
data
classical
RL
is
suitable
multi-task
recommendations
online
platforms.
To
address
these
challenges,
proposed
framework
integrates
Deep
(DRL)
multi-agent
approach.
This
adaptive
system
personalizes
experience
by
considering
key
factors
learner
sentiments,
style,
preferences,
competency,
difficulty
levels.
We
formulate
interactive
problem
using
DRL-based
Actor-Critic
model
named
DRR,
treating
sequential
decision-making
process.
The
DRR
enables
provide
top-N
paths,
enriching
student's
experience.
Extensive
experiments
MOOC
dataset
100
K
Coursera
review
validate
model,
demonstrating
its
superiority
over
baseline
models
major
evaluation
metrics
long-term
recommendations.
outcomes
research
contribute
field
e-learning
guiding
design
implementation
RSs,
facilitate
relevant
students.
ACM Transactions on Software Engineering and Methodology,
Journal Year:
2025,
Volume and Issue:
unknown
Published: March 7, 2025
Deep
reinforcement
learning
(DRL)
systems
have
been
increasingly
applied
in
various
domains.
Testing
them,
however,
remains
a
major
open
research
problem.
Mutation
testing
is
popular
test
suite
evaluation
technique
that
analyzes
the
extent
to
which
suites
detect
injected
faults.
It
has
widely
researched
both
traditional
software
and
field
of
deep
learning.
However,
due
fundamental
differences
between
DRL
software,
as
well
systems,
aspects
such
environment
interaction,
network
decision-making,
data
efficiency,
previous
mutation
techniques
cannot
be
directly
systems.
In
this
paper,
we
proposed
comprehensive
framework
specifically
designed
for
DRLMutation
,
further
fill
gap.
We
first
considered
characteristics
DRL,
based
on
training
process
model
trained
agent,
examined
combinations
from
three
dimensions:
objects,
operation
methods,
injection
methods.
This
approach
led
more
design
methodology
operators.
After
filtering,
identified
total
107
applicable
Then,
realm
evaluation,
formulated
set
metrics
tailored
assess
suites.
Finally,
validated
stealthiness
effectiveness
operators
Cart
Pole
Mountain
Car
Continuous
Lunar
Lander
Breakout
CARLA
environments.
show
inspiring
findings
majority
these
potentially
undermine
decision-making
capabilities
agent
without
affecting
normal
training.
The
varying
degrees
disruption
achieved
by
can
used
quality
different
