Thanks to visit codestin.com
Credit goes to www.scribd.com

Scribd
Upload
31 views7 pages

Soft Computing 300

Soft computing techniques like adaptive linear neurons (ADALINES) can be used to learn and model complex systems. ADALINES use a single layer neural network architecture with inputs, weights, bias and an output activation function. They are trained using the delta rule or Widrow-Hoff learning algorithm to minimize error between actual and target outputs by adjusting the weights and bias. Testing involves applying new input patterns, computing the weighted sum, applying the output activation function and checking the response. ADALINES can be designed to learn different logic functions by training on input-output pairs using the delta rule learning algorithm.

Uploaded by

rohit Singh
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
31 views7 pages

Soft Computing 300

Soft computing techniques like adaptive linear neurons (ADALINES) can be used to learn and model complex systems. ADALINES use a single layer neural network architecture with inputs, weights, bias and an output activation function. They are trained using the delta rule or Widrow-Hoff learning algorithm to minimize error between actual and target outputs by adjusting the weights and bias. Testing involves applying new input patterns, computing the weighted sum, applying the output activation function and checking the response. ADALINES can be designed to learn different logic functions by training on input-output pairs using the delta rule learning algorithm.

Uploaded by

rohit Singh
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 7

Soft Computing

300 (3)
Dr. Lekshmi R. R.
Asst. Prof.
Department of Electrical & Electronics
Engineering
Amrita School of Engineering 1
Adaptive Linear Neuron
(ADALINE)
Architecture
1
𝑏 Single layer neural network
𝑥1 𝑋1 𝑤1
y
𝑤2 𝑌
𝑥2
𝑋2

Single layer neural network


Training Identity 𝑦 = 𝑦𝑖𝑛
Testing 1 𝑖𝑓 𝑦𝑖𝑛 ≥ 0
Bipolar 𝑦 = ቊ
−1𝑖𝑓 𝑦𝑖𝑛 < 0
Training Algorithm
Delta/ Widrow Hoff rule
Step 1: Initialize weights and bias. Step 7: Adjust the weights and bias if error exist
Set learning rate α (0< α<1) (y≠t) new

Step 2: While termination condition is false, do 𝑤𝑖 = 𝑤𝑖 𝑜𝑙𝑑 + 𝛼(𝑡 − 𝑦)𝑥𝑖


steps 3 - 8 𝑏new = 𝑏𝑜𝑙𝑑 + 𝛼(𝑡 − 𝑦)
Step 3: For each training pair, do steps 4-7
Step 4: Set input activation functions; xi=si Else new
𝑤𝑖 = 𝑤𝑖 𝑜𝑙𝑑
Step 5: Compute response of system
𝑛 𝑏new = 𝑏𝑜𝑙𝑑
𝑦𝑖𝑛 = 𝑏 + ෍ 𝑥𝑖 𝑤𝑖
𝑖=1
Step 8: Test stopping condition
Step 6: Set output activation functions
𝑦 = 𝑦𝑖𝑛 – If no weights are changed, stop
– Else continue
Testing Algorithm
Step 1: Initialize weights and bias.
Step 2: For each training pair, do steps 3-5
Step 3: Set input activation functions; xi=si
Step 4: Compute response of system
𝑛

𝑦𝑖𝑛 = 𝑏 + ෍ 𝑥𝑖 𝑤𝑖
𝑖=1
Step 5: Set output activation functions
1 𝑖𝑓 𝑦𝑖𝑛 ≥ 0
𝑦=ቊ
−1 𝑖𝑓 𝑦𝑖𝑛 < 0
Design an ADALINE to implement OR gate
No. of inputs: 3 Weight Value
x1 x2 1 t
𝑤1 0.1
1 1 1 1 No. of outputs:1
𝑤2 0.2
1 -1 1 1
b 0.3
-1 1 1 1
-1 -1 1 -1 𝐧𝐞𝐰
Epoch 1 𝒘𝒊 = 𝒘𝒊 𝒐𝒍𝒅 + 𝜶(𝒕 − 𝒚)𝒙𝒊
𝒃𝐧𝐞𝐰 = 𝒃𝒐𝒍𝒅 + 𝜶(𝒕 − 𝒚)
α= 0.1 𝑦 = 𝑦𝑖𝑛

x1 x2 1 t yin y Y=t? Δw1 Δw2 Δb w1 w2 b


0.1 0.2 0.3
1 1 1 1 0.6 0.6 N 0.04 0.04 0.04 0.14 0.24 0.34

1 -1 1 1 0.24 0.24 N 0.076 -0.076 0.076 0.216 0.164 0.416

-1 1 1 1 0.364 0.364 N -0.0636 0.0636 0.0636 0.1524 0.2276 0.4796

-1 -1 1 -1 0.0996 0.0996 N 0.10996-0.10996-0.10996 0.263 0.3375 0.3696


Thank you

You might also like