Molecular insights into diabetic wound healing: Focus on Wnt/β-catenin and MAPK/ERK signaling pathways
Cytokine,
Год журнала:
2025,
Номер
191, С. 156957 - 156957
Опубликована: Май 13, 2025
Язык: Английский
Advancing neural computation: experimental validation and optimization of dendritic learning in feedforward tree networks
American Journal of Neurodegenerative Disease,
Год журнала:
2024,
Номер
13(5), С. 49 - 69
Опубликована: Янв. 1, 2024
This
study
aims
to
explore
the
capabilities
of
dendritic
learning
within
feedforward
tree
networks
(FFTN)
in
comparison
traditional
synaptic
plasticity
models,
particularly
context
digit
recognition
tasks
using
MNIST
dataset.
We
employed
FFTNs
with
nonlinear
segment
amplification
and
Hebbian
rules
enhance
computational
efficiency.
The
dataset,
consisting
70,000
images
handwritten
digits,
was
used
for
training
testing.
Key
performance
metrics,
including
accuracy,
precision,
recall,
F1-score,
were
analysed.
models
significantly
outperformed
plasticity-based
across
all
metrics.
Specifically,
framework
achieved
a
test
accuracy
91%,
compared
88%
demonstrating
superior
classification.
Dendritic
offers
more
powerful
by
closely
mimicking
biological
neural
processes,
providing
enhanced
efficiency
scalability.
These
findings
have
important
implications
advancing
both
artificial
intelligence
systems
neuroscience.
Язык: Английский
Quantized Context Based LIF Neurons for Recurrent Spiking Neural Networks in 45nm
Опубликована: Апрель 23, 2024
In
this
study,
we
propose
the
first
hardware
implementation
of
a
context-based
recurrent
spiking
neural
network
(RSNN)
emphasizing
on
integrating
dual
information
streams
within
neocortical
pyramidal
neurons
specifically
Context-Dependent
Leaky
Integrate
and
Fire
(CLIF)
neuron
models,
essential
element
in
RSNN.
We
present
quantized
version
CLIF
(qCLIF),
developed
through
hardware-software
codesign
approach
utilizing
sparse
activity
Implemented
45nm
technology
node,
qCLIF
is
compact
(900um²)
achieves
high
accuracy
90%
despite
8
bit
quantization
DVS
gesture
classification
dataset.
Our
analysis
spans
configuration
from
10
to
200
neurons,
supporting
up
82k
synapses
1.86
mm²
footprint,
demonstrating
scalability
efficiency.
Язык: Английский