|
Τ
|
Past observation length in terms of number of slot(s)
|
|
|
Output of the neuron of the k th layer
|
|
v
βα
|
Connection weights, connecting the neuron β of the k th layer to the neuron α in the (k −1)th layer
|
|
|
Weighted sum of inputs coming from the output neuron in the (k − 1) th layer
|
|
b
β
|
Bias input in the neuron
|
|
O
|
Neuron in the output layer
|
|
s
τ + 1
|
The actual slot(s) state from the MLP-based CSIT predictor
|
|
|
The desired slot(s) state
|
|
i
τ + 1
|
Actual predictor idle time slot(s)
|
|
|
Desired idle time slot(s)
|
|
φ'()
|
Activation function, i.e., ‘purelin’ for the output layer and ‘log sigmoid’ for the hidden layer
|
|
|
Local gradient of the neuron β in the k th layer
|
|
e
s
|
Error between the desired and actual slot(s) state of the predictor
|
|
e
i
|
Error between the desired and the actual idle time slot(s) of the predictor
|
|
N
is
|
Total number of idle time slot(s) in the system
|
|
N
i
|
Total number of idle time slot(s) sensed by CUsense
|
|
N
ip
|
Total number of idle time slot(s) sensed by CUpredict
|
|
X
|
Unit energy required for sensing a slot(s)
|
|
Y
|
Total number of slot(s) predicted to be idle
|
|
Z
|
Number of slots a CUsense sensed in a finite duration of time slot(s)
|
|
N
s
|
Total number of slot(s) to be sensed
|
|
C
|
Sensing threshold for the event PLR to occur
|
|
p
m
|
Probability of the m th slot(s) appearing to be idle
|
|
Δ
s
|
Time duration of the slot(s)
|
|
C
0
|
Referred transmission capacity of the SU
|
|
|
This corresponds to how fast the a SU can find the idle slot(s) among total of Ns
|
