IEEE Transactions on Pattern Analysis and Machine Intelligence,
Год журнала:
2015,
Номер
37(10), С. 2028 - 2040
Опубликована: Янв. 15, 2015
This
paper
introduces
a
spiking
hierarchical
model
for
object
recognition
which
utilizes
the
precise
timing
information
inherently
present
in
output
of
biologically
inspired
asynchronous
Address
Event
Representation
(AER)
vision
sensors.
The
nature
these
systems
frees
computation
and
communication
from
rigid
predetermined
enforced
by
system
clocks
conventional
systems.
Freedom
constraints
opens
possibility
using
true
to
our
advantage
computation.
We
show
not
only
how
can
be
used
recognition,
but
also
it
fact
simplify
Specifically,
we
rely
on
simple
temporal-winner-take-all
rather
than
more
computationally
intensive
synchronous
operations
typically
neural
networks
recognition.
approach
visual
represents
major
paradigm
shift
clocked
find
application
other
sensory
modalities
computational
tasks.
showcase
effectiveness
achieving
highest
reported
accuracy
date
(97.5\%$\pm$3.5\%)
previously
published
four
class
card
pip
task
an
84.9\%$\pm$1.9\%
new
difficult
36
character
task.
PLoS Computational Biology,
Год журнала:
2014,
Номер
10(4), С. e1003553 - e1003553
Опубликована: Апрель 17, 2014
Neuronal
population
codes
are
increasingly
being
investigated
with
multivariate
pattern-information
analyses.
A
key
challenge
is
to
use
measured
brain-activity
patterns
test
computational
models
of
brain
information
processing.
One
approach
this
problem
representational
similarity
analysis
(RSA),
which
characterizes
a
representation
in
or
model
by
the
distance
matrix
response
elicited
set
stimuli.
The
encapsulates
what
distinctions
between
stimuli
emphasized
and
de-emphasized
representation.
tested
comparing
it
predicts
that
region.
RSA
also
enables
us
compare
representations
stages
processing
within
given
model,
behavioral
data,
individuals
species.
Here,
we
introduce
Matlab
toolbox
for
RSA.
supports
an
simultaneously
data-
hypothesis-driven.
It
designed
help
integrate
wide
range
into
multichannel
measurements
as
provided
modern
functional
imaging
neuronal
recording
techniques.
Tools
visualization
inference
enable
user
relate
sets
regions
statistically
using
nonparametric
methods.
searchlight-based
RSA,
continuously
map
volume
search
code
specific
geometry.
Finally,
linear-discriminant
t
value
measure
discriminability
bridges
gap
linear
decoding
analyses
In
order
demonstrate
capabilities
toolbox,
apply
both
simulated
real
fMRI
data.
functions
equally
applicable
other
modalities
measurement.
freely
available
community
under
open-source
license
agreement
(http://www.mrc-cbu.cam.ac.uk/methods-and-resources/toolboxes/license/).
Chemical Reviews,
Год журнала:
2015,
Номер
116(1), С. 215 - 257
Опубликована: Дек. 21, 2015
Nano-bioelectronics
represents
a
rapidly
expanding
interdisciplinary
field
that
combines
nanomaterials
with
biology
and
electronics
and,
in
so
doing,
offers
the
potential
to
overcome
existing
challenges
bioelectronics.
In
particular,
shrinking
electronic
transducer
dimensions
nanoscale
making
their
properties
appear
more
biological
can
yield
significant
improvements
sensitivity
biocompatibility
thereby
open
up
opportunities
fundamental
healthcare.
This
review
emphasizes
recent
advances
nano-bioelectronics
enabled
semiconductor
nanostructures,
including
silicon
nanowires,
carbon
nanotubes,
graphene.
First,
synthesis
electrical
of
these
are
discussed
context
Second,
affinity-based
nano-bioelectronic
sensors
for
highly
sensitive
analysis
biomolecules
reviewed.
studies,
nanostructures
as
transistor-based
biosensors
from
device
behavior
through
sensing
applications
future
challenges.
Third,
complex
interface
between
nanoelectronics
living
systems,
single
cells
live
animals,
is
discussion
focuses
on
representative
electrophysiology
using
nanoelectronic
devices
cellular
measurements
emerging
work
where
arrays
incorporated
within
three-dimensional
cell
networks
define
synthetic
natural
tissues.
Last,
some
exciting
discussed.