Respiratory Motion Estimation with Hybrid Implementation - Prediction and Classification of Respiratory Motion - page 78

anywhere from 5 min to 1.5 h of the average time at the Georgetown University
Cyberknife treatment facility. These records were arbitrarily selected to represent a
wide variety of breathing patterns, including highly unstable and irregular
examples. Each patient's breathing record was used to independently train and test
the predictive accuracy of the filter.

4.4.2 Optimized Group Number for RMLP

(1) Optimized Group Number

With the respect to the selected group number (g) to implement the RMLP, we
used a multilayer perceptron with two hidden layers, where the first hidden layer is
recurrent and the second one is not. We increased the number of hidden units for
the first and the second hidden layer according to the group number to calculate
the objective function value for comparing two different methods. In order to
analyze the group number for RMLP, we incorporated objective function ( 4.16 )in
Sect. 4.3.3 .
As shown in Fig. 4.6 , HEKF is optimized when the group number is 2, whereas
DEKF is optimized when the group number is 6. Therefore, we choose the neuron
number 2 for HEKF and 6 for DEKF.

6

HEKF
DEKF

5

4

3

2

1

0

1 2 3 4 5 6 7 8 9

Neuron Numbers for each hidden layer (g)

Fig. 4.6 Comparison of objective function values between HEKF and DEKF. With this figure,
we can expect to choose the selected neuron number for HEKF or DEKF to be more optimized.
Also, the discriminant criterion itself tests whether HEKF or DEKF is less enormous