ANN Session−2
Example−2:
3 NN :
Input layer :3 neurons
Hidden layer :2 neurons
Output layer :1neurons
Input layer outputs are=Inputs only
Hidden layer inputs are x 1 , x 2 , x 3 also w 11 , w12 , w 21 , w 22 , w31 , w32
o 1=Activation(bh 1 +w 11∗x 1 +w 21∗x 2+ w31∗x 3)
o 2= Activation(bh 2 +w 12∗x1 + w22∗x 2 +w 32∗x3 )
How many weights at hiddenlayer : 6
number of neurons∈input layer∗number of neurons ∈hiddenlayer
How many parameters at hiddenlayer=weights+bias=6+2=8
number of neurons∈input layer∗number of neurons ∈hiddenlayer+ number of neurons ∈hiddenlayer
¿
o 1=Activation(bh 1 +w 11∗x 1 +w 21∗x 2+ w31∗x 3)
o 2= Activation(bh 1 +w 12∗x1 + w22∗x 2 +w32∗x 3 )
Output layer
y= Activation(b 0 +w h11∗o 1+ wh 12∗o2)
How many weights at hiddenlayer : 2
number of neurons∈hiddenlayer∗number of neurons∈ output layer=2∗1=2
How many parameters at hiddenlayer=weights+bias=2+1=3
number of neurons∈hiddenlayer layer∗number of neurons∈output layer+ number of neurons∈output layer
Total parameters∈above network=8+3=11
Maths Calculations
at Input layer :
No bias−No weights :Total parameters=0
At Hidden Layer :
Hidden layer−1 :Neuron ( 1 ) :b h 1+ w11∗x 1+ w21∗x 2 +w 31∗x 3
Hidden layer−1 :Neuron ( 2 ) : bh 2+ w12∗x 1 +w 22∗x 2+ w32∗x 3 ¿
Bias ( 2 ) :bh 1 , bh 2 ,Weihgts ( 6 ) : w11 , w 21 , w12 , w22 , w31 , w32
Total parameters=2+6=8
Trick : Hiddenlayer parameters
inputlaye r neurons∗Hiddenlaye r Neurons + Hiddenlaye r Neurons
inputlaye r neurons∗Hiddenlaye r Neurons : Number of weights
Hiddenlaye r Neurons : Number of Bias
At Output layer :
o 1=Activation ( wh 1+ w11∗x 1 +w 21∗x2 )
o 2= Activation ( wh 2 +w 12∗x1 + w22∗x 2 )
y= Activation(wo + wh 1 o∗o1+ w h2 o∗o 2)
Trick :output layer parameters
Hiddenlaye r Neurons∗outpu t layer +outpu t layer
2∗1+ 1=3
4 NN
Input layer=3 neurons
Hidden layer−1 :3 neurons
Hidden layer−2 :2 neurons
output layer :1 neuron
layer (nuerons) Parameter (weights+bias)
Input layer (3) 0
Hidden layer (3) 3∗3+3=12
Hidden layer (2) 3∗2+2=8
output layer (1) 2∗1+ 1=3
23
5NN
Input layer=10 neurons
Hidden layer−1 :10 neurons
Hidden layer−2 :5 neurons
Hidden layer−2 :3 neurons
output layer :10 neuron
Total :183 185 216 , 222 , 223 ,282 , 183
layer (nuerons) Parameter (weights+bias)
Input layer (10) 0
Hidden layer (10) 10∗10+10=110
Hidden layer (5) 10∗5+5=55
Hidden layer (3) 5∗3+3=18
output layer (10) 3∗10+10=40
223
6 NN
Input layer=20 neurons
Hidden layer−1 :15 neurons
Hidden layer−2 :8 neurons
Hidden layer−3 :5 neurons
Hidden layer−4 :3 neuron
output layer :10 neuron
Type equation here .
layer (nuerons) Parameter (weights+bias)
Input layer (20) 0
Hidden layer (15) 20∗15+15=315
Hidden layer (8) 15∗8+8=128
Hidden layer (5) 8∗5+5=45
Hidden layer (3) 5∗3+3=18
output layer (10) 3∗10+10=40
546
1)
2)
How ¿ find the parameters
3)
what is the activation function
4)
How ¿ choose the hiddenlayers
5)
How ¿ avoid the overfitting
How ¿ speed up the process
