Basic Neural Network
- Sunny Chandra
©c17hawke
©c17hawke
INPUT
ARCHITECTURE OF SINGLE NEURON
w_{11}
w_{12}
\sum \rightarrow
\\
z = w_{11}\times x_1 + w_{12}\times x_2 + b
act(z)
out = act(z)
input\\ x_1
input\\ x_2
bias\\ b
©c17hawke
AN EXAMPLE OF ACTIVATION FUNCTION
©c17hawke
1
2
3
1
2
3
INPUT
OUTPUT
Input layer ( buffer layer )
1
2
3
©c17hawke
INPUT
OUTPUT
THIS CAN BE AN EXAMPLE OF CLASSIFICATION NN WITH 3 DIFFERENT CLASS
1
2
3
1
2
3
1
2
3
©c17hawke
1
INPUT
OUTPUT
REGRESSION NN EXAMPLE
1
2
3
1
2
3
©c17hawke
1
2
1
2
w_{11}
w_{12}
w_{21}
w_{22}
z_1 = (w_{11} \times i_1) + (w_{12} \times i_2)
z_2 = (w_{21} \times i_1) + (w_{22} \times i_2)
©c17hawke
=
=
=
=
*
(w_{11} \times i_1) + (w_{12} \times i_2)
(w_{21} \times i_1) + (w_{22} \times i_2)
©c17hawke
=
=
=
=
*
(w_{11} \times i_1) + (w_{12} \times i_2)
(w_{21} \times i_1) + (w_{22} \times i_2)
W
©c17hawke
=
=
=
=
*
(w_{11} \times i_1) + (w_{12} \times i_2)
(w_{21} \times i_1) + (w_{22} \times i_2)
W
W
X
©c17hawke
=
=
=
=
*
(w_{11} \times i_1) + (w_{12} \times i_2)
(w_{21} \times i_1) + (w_{22} \times i_2)
W
W
X
Z
©c17hawke
=
=
=
=
*
(w_{11} \times i_1) + (w_{12} \times i_2)
(w_{21} \times i_1) + (w_{22} \times i_2)
W
W
X
Z
act(Z)
©c17hawke
=
=
=
=
*
(w_{11} \times i_1) + (w_{12} \times i_2)
(w_{21} \times i_1) + (w_{22} \times i_2)
W
W
X
Z
act(Z)
\hat{y}
©c17hawke
=
=
=
*
W
W
X
Z
act(Z)
\hat{y}
©c17hawke
1
2
1
2
predicted val
actual val
error
ERROR CALCULATION
(\ \ \ \ \ \ \ \ \ \ )^2
(\ \ \ \ \ \ \ \ \ \ )^2
©c17hawke
1
2
1
error
ERROR CALCULATION
= 3.0
= 1.0
2
©c17hawke
1
2
1
error
ERROR CALCULATION
= 3.0
= 1.0
1/2 error ?
1/2 error ?
2
©c17hawke
1
2
1
error
ERROR CALCULATION
An intuition
= 3.0
= 1.0
3/4 error !!
1/4 error !!
2
©c17hawke