Biomedical Engineering Reference
In-Depth Information
Layer 1
Layer 2
Layer 3
Layer 4
Layer 5
μ
A1
(
x
)
A
1
w
1
ŵ
1
N
Π
x
ŵ
1
f
1
μ
A2
(
x
)
A
2
x y
f
Σ
B
1
μ
B1
(
y
)
ŵ
2
f
2
y
Π
N
B
2
w
2
ŵ
2
μ
B2
(
y
)
Fig. 2.13 Adaptive neuro fuzzy inference system with the total five layers. Based on the
incoming elements (x and y), this system is composed with two layers of adaptive nodes (layers 1
and 4) and three layers of fixed nodes (layers 2, 3, and 5). The layer 1 is characterized by a
membership function l(
) that assigns each incoming element to a value between 0 and 1. Layer 4
is trained by a least squares method
We summarized the prediction accuracy and a representative feature for each
method of the model-free approach, as shown in Table
2.3
.
2.3.3 Hybrid Prediction Algorithms
Hybrid prediction algorithms used united methods to combine more than two
methods or approaches to obtain outstanding results, compared to a previous
solitary method. This method includes (1) adaptive neuro-fuzzy interference sys-
tem (ANFIS) [
80
,
97
], (2) hybrid model with adaptive filter and nonlinear model
(Adaptive Tumor Tracking System) [
68
,
98
], and (3) interacting multiple model
(IMM) filter [
72
,
81
,
83
].
2.3.3.1 Adaptive Neuro-Fuzzy Inference System
A adaptive neuro-fuzzy inference system (ANFIS) is a hybrid intelligent system
with combining both learning capabilities of a neural network and fuzzy logic
reasoning, to find a specific model in association with input breathing motion and
target prediction. The proposed neuro-fuzzy model ANFIS in [
80
] is a multilayer
neural network-based fuzzy system in combination with two layers of adaptive
nodes (layer 1 and 4) and three layers of fixed nodes (layer 2, 3, and 5), as shown in
Fig.
2.13
[
80
].
The first layer is distinguished by a fuzzy set (A
1
,A
2
,B
1
,B
2
) that is expressed
by a membership function to assign each incoming element to a membership value
between 0 and 1, as the following equation:
Search WWH ::
Custom Search