Biomedical Engineering Reference
In-Depth Information
H
Set of finite mixture model parameter vectors
i.e. H
a
y
;
m
y
;
R
y
G
y
¼
1
pz;
ð Þ
Joint probability density that consists of the mixture of Gaussians
py
j
z
j
Posterior probability value of group y with z
j
d
ðÞ
Log-likelihood function with G components
D(G)
Difference of the consecutive log-likelihood functions
b
y
Hyper-parameter that presents some background knowledge as a
hypothetical prior probability
b
y
(k)
Adaptive hyper-parameter
Dl
y
Difference between the current channel selection probability and
the previous one in the group y
d
ADT
Log-likelihood function with the adaptive posterior probability
l
ab
Channel
selection
probability
that
represents
the
conditional
transition probability from channel a to channel b
T(k)
Asymptotic lower bound of recursive computation based on time k
T(L)
Lower bound of iteration execution time for k-means clustering
based on L points
u
Input vector with external and feedback inputs
w
Weights
v
Internal activation function of a neuron
U
Nonlinear activation function
y
i
Output of the ith neuron
x(k)
External input of a system model at time k
y(k)
Output of a system model at time k
ðÞ
T
Vector transpose operator
w(k)
Weight vector of the entire network at time k
D(k)
Desired (teaching) signal at time k
s
Number of weights in the entire network
p
Number of output nodes
v(k)
Recurrent activities inside the network at time k
u(k)
Input signal applied to the network at time k
Q(k)
Process noise with the property of a multivariate zero-mean white
noise
r(k)
Measurement noise with the property of a multivariate zero-mean
white noise
b
;
;
ð
Þ
Measurement function that accounts for the overall nonlinearity of
the multilayer perceptron from the input to the output layer
B(k)
p 9 s measurement matrix of the linearized model
(k)
p 9 1 matrix denoting the difference between the desired response
d(k) and its estimation
^wkk
1
ð
j
Þ
s 9 1 vector denoting the estimate of the weight vector w(k) at
time k given the observed data up to time k-1
ðÞ
(=
ˆ
(k+1|k)) Filtered updated estimate of w(k) on receipt of the
observable d(k)
^wk
Search WWH ::
Custom Search