NeuroImage,
Год журнала:
2016,
Номер
145, С. 180 - 199
Опубликована: Июнь 28, 2016
Neuroimaging
increasingly
exploits
machine
learning
techniques
in
an
attempt
to
achieve
clinically
relevant
single-subject
predictions.
An
alternative
learning,
which
tries
establish
predictive
links
between
features
of
the
observed
data
and
clinical
variables,
is
deployment
computational
models
for
inferring
on
(patho)physiological
cognitive
mechanisms
that
generate
behavioural
neuroimaging
responses.
This
paper
discusses
rationale
behind
a
approach
neuroimaging-based
inference,
focusing
its
potential
characterising
disease
individual
subjects
mapping
these
characterisations
Following
overview
two
main
approaches
–
Bayesian
model
selection
generative
embedding
can
link
predictions,
we
review
how
methods
accommodate
heterogeneity
psychiatric
neurological
spectrum
disorders,
help
avoid
erroneous
interpretations
data,
mechanistic,
model-based
statistical
perspectives
afforded
by
learning.
Neuropsychopharmacology,
Год журнала:
2021,
Номер
47(1), С. 72 - 89
Опубликована: Авг. 18, 2021
Concepts
of
cognitive
control
(CC)
and
executive
function
(EF)
are
defined
in
terms
their
relationships
with
goal-directed
behavior
versus
habits
controlled
automatic
processing,
related
to
the
functions
prefrontal
cortex
(PFC)
regions
networks.
A
psychometric
approach
shows
unity
diversity
CC
constructs,
3
components
most
commonly
studied
constructs:
general
or
common
specific
mental
set
shifting
working
memory
updating.
These
constructs
considered
against
cellular
systems
neurobiology
PFC
what
is
known
its
functional
neuroanatomical
network
organization
based
on
lesioning,
neurochemical,
neuroimaging
approaches
across
species.
also
context
motivation,
as
"cool"
"hot"
forms.
Its
Common
component
shown
be
distinct
from
intelligence
(g)
closely
response
inhibition.
Impairments
possible
causes
psychiatric
symptoms
consequences
disorders.
The
factor
psychopathology
(p)
dimensional
such
impulsivity
large
scale
developmental
adult
populations
considered,
well
implications
for
genetic
studies
RDoC
classification.
Computational
modeling
of
behavior
has
revolutionized
psychology
and
neuroscience.
By
fitting
models
to
experimental
data
we
can
probe
the
algorithms
underlying
behavior,
find
neural
correlates
computational
variables
better
understand
effects
drugs,
illness
interventions.
But
with
great
power
comes
responsibility.
Here,
offer
ten
simple
rules
ensure
that
is
used
care
yields
meaningful
insights.
In
particular,
present
a
beginner-friendly,
pragmatic
details-oriented
introduction
on
how
relate
data.
What,
exactly,
model
tell
us
about
mind?
To
answer
this,
apply
our
simplest
techniques
most
accessible
beginning
modelers
illustrate
them
examples
code
available
online.
However,
more
advanced
techniques.
Our
hope
by
following
guidelines,
researchers
will
avoid
many
pitfalls
unleash
their
own
Annual Review of Psychology,
Год журнала:
2016,
Номер
68(1), С. 101 - 128
Опубликована: Сен. 13, 2016
We
review
the
psychology
and
neuroscience
of
reinforcement
learning
(RL),
which
has
experienced
significant
progress
in
past
two
decades,
enabled
by
comprehensive
experimental
study
simple
decision-making
tasks.
However,
one
challenge
RL
is
computational:
The
simplicity
these
tasks
ignores
important
aspects
real
world:
(a)
State
spaces
are
high-dimensional,
continuous,
partially
observable;
this
implies
that
(b)
data
relatively
sparse
and,
indeed,
precisely
same
situation
may
never
be
encountered
twice;
furthermore,
(c)
rewards
depend
on
long-term
consequences
actions
ways
violate
classical
assumptions
make
tractable.
A
seemingly
distinct
that,
cognitively,
theories
have
largely
involved
procedural
semantic
memory,
way
knowledge
about
action
values
or
world
models
extracted
gradually
from
many
experiences
can
drive
choice.
This
focus
memory
leaves
out
such
as
episodic
related
to
traces
individual
events.
suggest
challenges
related.
computational
dealt
with,
part,
endowing
systems
with
allowing
them
efficiently
approximate
value
functions
over
complex
state
spaces,
learn
very
little
data,
bridge
dependencies
between
rewards.
theory
underlying
proposal
empirical
evidence
support
it.
Our
suggests
ubiquitous
diverse
roles
function
part
an
integrated
system.
Clinical Psychological Science,
Год журнала:
2017,
Номер
5(5), С. 880 - 889
Опубликована: Июнь 11, 2017
Evidence
indicates
the
existence
of
a
superordinate
factor
general
psychopathology,
which
has
been
termed
p.
Among
issues
raised
by
this
discovery
is
whether
substantive
meaning
or
not.
This
article
suggests
functional
interpretation
p
factor,
based
in
part
on
family
dual
process
models,
an
associative
system
and
deliberative
compete
for
influence
over
action.
The
frequently
said
to
be
impulsively
responsive
emotions.
We
hypothesize
that
impulsive
responsivity
emotion
underlies
factor.
One
benefit
view
use
same
underlying
variable
account
both
internalizing
externalizing
vulnerabilities
as
well
aspects
thought
disorder.
reviewed
linking
reactivity
(separately)
internalizing,
externalizing,
thought-disorder
symptoms.
Alternative
interpretations
are
considered.
Psychological Review,
Год журнала:
2019,
Номер
126(2), С. 292 - 311
Опубликована: Янв. 24, 2019
Habits
form
a
crucial
component
of
behavior.In
recent
years,
key
computational
models
have
conceptualized
habits
as
arising
from
model-free
reinforcement
learning
(RL)
mechanisms,
which
typically
select
between
available
actions
based
on
the
future
value
expected
to
result
each.Traditionally,
however,
been
understood
behaviors
that
can
be
triggered
directly
by
stimulus,
without
requiring
animal
evaluate
outcomes.Here,
we
develop
model
instantiating
this
traditional
view,
in
through
direct
strengthening
recently
taken
rather
than
encoding
outcomes.We
demonstrate
accounts
for
behavioral
manifestations
habits,
including
insensitivity
outcome
devaluation
and
contingency
degradation,
well
effects
schedule
rate
habit
formation.The
also
explains
prevalent
observation
perseveration
repeated-choice
tasks
an
additional
manifestation
system.We
suggest
mapping
habitual
onto
value-free
mechanisms
provides
parsimonious
account
existing
neural
data.This
may
provide
new
foundation
building
robust
comprehensive
interaction
with
other,
more
goal-directed
types
help
better
guide
research
into
underlying
control
instrumental
behavior
generally.
Psychological Science,
Год журнала:
2017,
Номер
28(9), С. 1321 - 1333
Опубликована: Июль 21, 2017
Human
behavior
is
sometimes
determined
by
habit
and
other
times
goal-directed
planning.
Modern
reinforcement-learning
theories
formalize
this
distinction
as
a
competition
between
computationally
cheap
but
inaccurate
model-free
system
that
gives
rise
to
habits
expensive
accurate
model-based
implements
It
unclear,
however,
how
people
choose
allocate
control
these
systems.
Here,
we
propose
arbitration
occurs
comparing
each
system’s
task-specific
costs
benefits.
To
investigate
proposal,
conducted
two
experiments
showing
increase
when
it
achieves
greater
accuracy
than
control,
especially
the
rewards
of
performance
are
amplified.
In
contrast,
they
insensitive
reward
amplification
yield
equivalent
accuracy.
This
suggests
humans
adaptively
balance
habitual
planned
action
through
on-line
cost-benefit
analysis.
Neuropsychopharmacology,
Год журнала:
2020,
Номер
45(5), С. 720 - 735
Опубликована: Янв. 6, 2020
Drug
addiction
may
be
a
goal-directed
choice
driven
by
excessive
drug
value
in
negative
affective
states,
habit
strong
stimulus-response
associations,
or
compulsion
insensitivity
to
costs
imposed
on
seeking.
Laboratory
animal
and
human
evidence
for
these
three
theories
is
evaluated.
Excessive
goal
theory
supported
dependence
severity
being
associated
with
greater
choice/economic
demand.
demonstrably
(driven
the
expected
of
drug)
can
augmented
stress/negative
mood
induction
withdrawal-effects
amplified
those
psychiatric
symptoms
use
coping
motives.
Furthermore,
confer
risk
dependence,
motives
mediate
this
risk.
Habit
has
weaker
support.
Habitual
behaviour
seen
drug-exposed
animals
often
does
not
occur
complex
decision
scenarios,
where
responding
rewarded,
so
unlikely
explain
most
addictive
conditions
apply.
studies
have
found
propensity
habitual
users
as
function
severity,
minority
that
explained
task
disengagement
producing
impaired
explicit
contingency
knowledge.
Compulsion
also
weak
The
persistence
punished
seeking
better
(evinced
association
economic
demand)
than
costs.
provided
discount
cost
severity.
These
data
suggest
primarily
under
affect,
less
compulsion.
Addiction
pathological
because
states
powerfully
increase
acutely
outweighing
abstinence
goals.