bioRxiv (Cold Spring Harbor Laboratory),
Год журнала:
2024,
Номер
unknown
Опубликована: Ноя. 28, 2024
Abstract
Real
world
choices
often
involve
balancing
decisions
that
are
optimized
for
the
short-vs.
long-term.
Here,
we
reason
apparently
sub-optimal
single
trial
in
macaques
may
fact
reflect
long-term,
strategic
planning.
We
demonstrate
freely
navigating
VR
sequentially
presented
targets
will
strategically
abort
offers,
forgoing
more
immediate
rewards
on
individual
trials
to
maximize
session-long
returns.
This
behavior
is
highly
specific
individual,
demonstrating
about
their
own
long-run
performance.
Reinforcement-learning
(RL)
models
suggest
this
algorithmically
supported
by
modular
actor-critic
networks
with
a
policy
module
not
only
optimizing
long-term
value
functions,
but
also
informed
of
state-action
values
allowing
rapid
optimization.
The
artificial
suggests
changes
matched
offer
ought
be
evident
as
soon
offers
made,
even
if
aborting
occurs
much
later.
confirm
prediction
units
and
population
dynamics
macaque
dorsolateral
prefrontal
cortex
(dlPFC),
parietal
area
7a
or
dorsomedial
superior
temporal
(MSTd),
upcoming
reward-maximizing
upon
presentation.
These
results
cast
dlPFC
specialized
module,
stand
contrast
recent
work
distributed
recurrent
nature
belief-networks.
Social
animals
live
in
groups
and
interact
volitionally
complex
ways.
However,
little
is
known
about
neural
responses
under
such
natural
conditions.
Here,
we
investigated
hippocampal
CA1
neurons
a
mixed-sex
group
of
five
to
10
freely
behaving
wild
Egyptian
fruit
bats
that
lived
continuously
laboratory-based
cave
formed
stable
social
network.
In-flight,
most
place
cells
were
socially
modulated
represented
the
identity
sex
conspecifics.
Upon
interactions,
specific
interaction
types.
During
active
observation,
encoded
bat’s
own
position
head
direction,
together
with
position,
multiple
Identity-coding
same
bat
across
contexts.
The
strength
coding
was
by
sex,
hierarchy,
affiliation.
Thus,
form
multidimensional
sociospatial
representation
world.
Hippocampus,
Год журнала:
2023,
Номер
33(5), С. 586 - 599
Опубликована: Апрель 11, 2023
Abstract
The
discovery
of
place
cells
and
head
direction
in
the
hippocampal
formation
freely
foraging
rodents
has
led
to
an
emphasis
its
role
encoding
allocentric
spatial
relationships.
In
contrast,
studies
head‐fixed
primates
have
additionally
found
representations
views.
We
review
recent
experiments
moving
monkeys
that
expand
upon
these
findings
show
postural
variables
such
as
eye/head
movements
strongly
influence
neural
activity
formation,
suggesting
function
hippocampus
depends
on
where
animal
looks.
interpret
results
light
humans
performing
challenging
navigation
tasks
which
suggest
depending
context,
serve
one
two
roles—gathering
information
about
structure
environment
(active
sensing)
or
externalizing
contents
internal
beliefs/deliberation
(embodied
cognition).
These
prompt
future
experimental
investigations
into
carried
by
signals
flowing
between
brain
regions
controlling
variables,
constitute
a
basis
for
updating
computational
theories
system
accommodate
movements.
Philosophical Transactions of the Royal Society B Biological Sciences,
Год журнала:
2023,
Номер
378(1886)
Опубликована: Авг. 7, 2023
A
key
computation
in
building
adaptive
internal
models
of
the
external
world
is
to
ascribe
sensory
signals
their
likely
cause(s),
a
process
causal
inference
(CI).
CI
well
studied
within
framework
two-alternative
forced-choice
tasks,
but
less
understood
cadre
naturalistic
action-perception
loops.
Here,
we
examine
disambiguating
retinal
motion
caused
by
self-
and/or
object-motion
during
closed-loop
navigation.
First,
derive
normative
account
specifying
how
observers
ought
intercept
hidden
and
moving
targets
given
belief
about
(i)
whether
was
target
moving,
(ii)
if
so,
with
what
velocity.
Next,
line
modelling
results,
show
that
humans
report
as
stationary
steer
towards
initial
rather
than
final
position
more
often
when
they
are
themselves
suggesting
putative
misattribution
self.
Further,
predict
should
misattribute
often:
passive
active
self-motion
(given
lack
an
efference
copy
informing
estimates
former),
presented
eccentrically
centrally
lateral
flow
vectors
larger
at
eccentric
locations
forward
self-motion).
Results
support
both
these
predictions.
Lastly,
analysis
eye
movements
that,
while
saccades
toward
were
largely
accurate
regardless
condition,
subsequent
gaze
pursuit
modulated
velocity
object-only
motion,
not
concurrent
object-
self-motion.
These
results
demonstrate
loops,
suggest
protracted
temporal
unfolding
computations
characterizing
CI.
This
article
part
theme
issue
'Decision
control
processes
multisensory
perception'.
We
do
not
understand
how
neural
nodes
operate
and
coordinate
within
the
recurrent
action-perception
loops
that
characterize
naturalistic
self-environment
interactions.
Here,
we
record
single-unit
spiking
activity
local
field
potentials
(LFPs)
simultaneously
from
dorsomedial
superior
temporal
area
(MSTd),
parietal
7a,
dorsolateral
prefrontal
cortex
(dlPFC)
as
monkeys
navigate
in
virtual
reality
to
‘catch
fireflies’.
This
task
requires
animals
actively
sample
a
closed-loop
environment
while
concurrently
computing
continuous
latent
variables:
(i)
distance
angle
travelled
(i.e.,
path
integration)
(ii)
memorized
firefly
location
hidden
spatial
goal).
observed
patterned
mixed
selectivity,
with
most
prominently
coding
for
variables,
sensorimotor
MSTd
often
eye
movements.
However,
even
traditionally
considered
sensory
MSTd)
tracked
demonstrating
integration
vector
of
goals.
Further,
global
encoding
profiles
unit-to-unit
coupling
noise
correlations)
suggested
functional
subnetwork
composed
by
dlPFC,
between
these
anatomy
would
suggest.
show
greater
more
animals’
gaze
position
was
indicative
ongoing
goal.
suggest
this
MSTd-dlPFC
reflects
monkeys’
natural
adaptive
strategy
wherein
they
continuously
toward
(invisible)
target.
Together,
results
highlight
distributed
nature
during
closed
fine-grain
subnetworks
may
be
dynamically
established
subserve
(embodied)
strategies.
Proceedings of the National Academy of Sciences,
Год журнала:
2024,
Номер
121(14)
Опубликована: Март 29, 2024
During
foraging
behavior,
action
values
are
persistently
encoded
in
neural
activity
and
updated
depending
on
the
history
of
choice
outcomes.
What
is
mechanism
for
value
maintenance
updating?
Here,
we
explore
two
contrasting
network
models:
synaptic
learning
versus
integration.
We
show
that
both
models
can
reproduce
extant
experimental
data,
but
they
yield
distinct
predictions
about
underlying
biological
circuits.
In
particular,
integrator
model
not
requires
reward
signals
mediated
by
pools
selective
alternatives
their
projections
aligned
with
linear
attractor
axes
valuation
system.
demonstrate
experimentally
observable
dynamical
signatures
feasible
perturbations
to
differentiate
scenarios,
suggesting
a
more
robust
candidate
mechanism.
Overall,
this
work
provides
modeling
framework
guide
future
research
probabilistic
foraging.
The
brain
may
have
evolved
a
modular
architecture
for
daily
tasks,
with
circuits
featuring
functionally
specialized
modules
that
match
the
task
structure.
We
hypothesize
this
enables
better
learning
and
generalization
than
architectures
less
modules.
To
test
this,
we
trained
reinforcement
agents
various
neural
on
naturalistic
navigation
task.
found
agent,
an
segregates
computations
of
state
representation,
value,
action
into
modules,
achieved
generalization.
Its
learned
representation
combines
prediction
observation,
weighted
by
their
relative
uncertainty,
akin
to
recursive
Bayesian
estimation.
This
agent’s
behavior
also
resembles
macaques’
more
closely.
Our
results
shed
light
possible
rationale
brain’s
modularity
suggest
artificial
systems
can
use
insight
from
neuroscience
improve
in
natural
tasks.
Progress in Neurobiology,
Год журнала:
2023,
Номер
231, С. 102537 - 102537
Опубликована: Окт. 12, 2023
Classical
neurophysiology
suggests
that
the
motor
cortex
(MI)
has
a
unique
role
in
action
control.
In
contrast,
this
review
presents
evidence
for
multiple
parieto-frontal
spinal
command
modules
can
bypass
MI.
Five
observations
support
modular
perspective:
(i)
statistics
of
cortical
connectivity
demonstrate
functionally-related
clusters
areas,
defining
functional
premotor,
cingulate,
and
parietal
cortices;
(ii)
different
corticospinal
pathways
originate
from
above
each
with
distinct
range
conduction
velocities;
(iii)
activation
time
module
varies
depending
on
task,
be
activated
simultaneously;
(iv)
architecture
direct
output
is
faster
less
metabolically
expensive
than
an
relies
MI,
given
slow
connections
between
MI
other
areas;
(v)
lesions
areas
composing
have
effects
Here
we
provide
examples
six
cortico-spinal
functions
they
subserve:
1)
arm
reaching,
tool
use
object
construction;
2)
spatial
navigation
locomotion;
3)
grasping
observation
hand
mouth
actions;
4)
initiation,
sequences,
encoding;
5)
conditional
association
learning,
plan
switching
inhibition;
6)
planning
defensive
actions.
These
serve
as
library
tools
to
recombined
when
faced
novel
tasks,
might
recombinatory
hub.
conclusion,
availability
locally-stored
information
outflow
paths
supports
physiological
plausibility
proposed
perspective.
Journal of Neuroscience,
Год журнала:
2022,
Номер
42(45), С. 8450 - 8459
Опубликована: Ноя. 9, 2022
Since
the
discovery
of
conspicuously
spatially
tuned
neurons
in
hippocampal
formation
over
50
years
ago,
characterizing
which,
where,
and
how
encode
navigationally
relevant
variables
has
been
a
major
thrust
navigational
neuroscience.
While
much
this
effort
centered
on
functionally-adjacent
structures,
recent
work
suggests
that
spatial
codes,
some
form
or
another,
can
be
found
throughout
brain,
even
areas
traditionally
associated
with
sensation,
movement,
executive
function.
In
review,
we
highlight
these
unexpected
results,
draw
insights
from
comparison
codes
across
contexts,
regions,
species,
finally
suggest
an
avenue
for
future
to
make
sense
diverse
dynamic
codes.