Scribd
Upload
47 views1 page

7.3.3.3 Empat-Lapisan Sistem Neuro-Fuzzy (ANFIS)

The document discusses two alternative architectures for neuro-fuzzy systems: a four-layer and three-layer structure. The four-layer neuro-fuzzy system (ANFIS) combines the third and fourth layers of a typical five-layer structure into a single layer for normalization. The three-layer neuro-fuzzy approximator uses a Sugeno-type model and represents each fuzzy rule with a singleton consequent. It consists of an input, hidden and output layer where the hidden layer has nodes based on the number of fuzzy sets per input.

Uploaded by

api-27091270
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as RTF, PDF, TXT or read online on Scribd
47 views1 page

7.3.3.3 Empat-Lapisan Sistem Neuro-Fuzzy (ANFIS)

The document discusses two alternative architectures for neuro-fuzzy systems: a four-layer and three-layer structure. The four-layer neuro-fuzzy system (ANFIS) combines the third and fourth layers of a typical five-layer structure into a single layer for normalization. The three-layer neuro-fuzzy approximator uses a Sugeno-type model and represents each fuzzy rule with a singleton consequent. It consists of an input, hidden and output layer where the hidden layer has nodes based on the number of fuzzy sets per input.

Uploaded by

api-27091270
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as RTF, PDF, TXT or read online on Scribd
You are on page 1/ 1

7.3.3.

3 four-layer neuro-fuzzy system (ANFIS)


although the five-layer structure described here has been so far the nost commonly known
for ANFIS architecture, there has been another suggested structure of ANFIS involving a four-layer
architecture. in this structure , layers three and four pertaining to normalization and consequent
variables are combined in such a way as to obtain an equivalent structure with only four layers. in
the new network the weight normalization takes place at the last layer[15]. figture 7.8 shows the
ANFIS four-layer structure with two inputs.

7.3.3.4 three-layer neuro-fuzzy approximator


in[22], L.maguire and co-authors introduced a three-layer neuro-fuzzy system as an
alternative architecture for approximate fuzzy reasoning. the fuzzy model used in their works is a
zero-order sugeno model, where each fuzzy rule has a euzzy singleton consequent. the proposed
neuro-fuzzy architecture implements the zero-order sugeno type as a three-layer network. the first
laye consist of n+1 nodes where n is the number of inputs. the extra node represents a unity bias
input. the input layaer (first layer) broadcasts the inputs to selected nodes in the hidden layer. the
number of nodes in the hidden layer depends on the number of fuzzy sets associated with each input
domain. the number of nodes in the hidden layer is determined by the relation np, where p is the
number of fuzzy sets (partitions) in each input domain and n is the number of inputs.
each node in the hidden layer receives two weighted inputs, one from one of the input
variables and the second from the bias node. in the hidden node, the squared sum of those variables
is passed to an activation function to produce an output according to:

terjemahan

7.3.3.3 empat-lapisan sistem neuro-fuzzy (ANFIS)

walaupun lima-lapis struktur yang dijelaskan di sini telah nost yang selama ini dikenal untuk
arsitektur ANFIS, ada saran lain struktur ANFIS melibatkan empat-lapisan arsitektur. dalam struktur
ini, lapisan tiga dan empat menyangkut hal menjadikan normal dan akibat variabel digabungkan
dalam sebaik mungkin untuk mendapatkan struktur yang setara dengan hanya empat lapisan.
jaringan baru dalam hal menjadikan normal berat berlangsung di lapisan terakhir [15]. figture 7,8
menunjukkan ANFIS empat-lapisan struktur dengan dua masukan.

7.3.3.4 tiga-lapisan neuro-fuzzy approximator

dalam [22], L.maguire dan co-penulis memperkenalkan tiga-lapisan neuro-fuzzy sistem sebagai
alternatif untuk arsitektur perkiraan fuzzy alasan. fuzzy model yang digunakan dalam karya mereka
adalah nol-pesanan sugeno model, fuzzy aturan di mana masing-masing memiliki akibat euzzy
tunggal. usulan neuro-fuzzy arsitektur melaksanakan nol-pesanan sugeno jenis sebagai tiga lapisan
jaringan. laye pertama terdiri dari node n +1 di mana n adalah jumlah input. ekstra node merupakan
kesatuan bias masukan. input layaer (lapisan pertama) siaran masukan yang dipilih untuk node di
lapisan tersembunyi. jumlah node dalam lapisan tersembunyi tergantung pada jumlah fuzzy set
yang terkait dengan setiap masukan domain. jumlah node dalam lapisan tersembunyi ditentukan
oleh hubungan np, dimana p adalah jumlah fuzzy set (partisi) di setiap domain masukan dan n
adalah jumlah input.

setiap node tersembunyi di lapisan dua weighted menerima masukan, satu dari salah satu variabel
dan masukan dari kedua bias node. tersembunyi di node, yang squared jumlah variabel yang dibawa
ke aktivasi fungsi untuk menghasilkan output yang sesuai untuk:

You might also like