Escolar Documentos
Profissional Documentos
Cultura Documentos
1
MLP
jth neuron
3
Over all samples
Function of all
free parameters
4
Online learning:
- Training is done example by example basis
- Cost function (optimization) instantaneous error energy
- Synaptic weights are adjusted on sample to sample basis
5
The learning curve is Final error value (vs) Epoch
6
Back Propagation
Online training for multilayer
perceptions
7
Induced local field / Total activation:
Activation function:
Cost function:
1 2
𝜀𝜀 𝝎𝝎(𝑛𝑛) = 𝒀𝒀(𝑛𝑛) − 𝑫𝑫(𝑛𝑛)
2
Basing on gradient decent
(similar to LMS): 𝜕𝜕𝜀𝜀(𝑛𝑛)
∆𝝎𝝎(𝑛𝑛) = −η
𝜕𝜕𝝎𝝎(𝑛𝑛)
8
Sensitivity factor (how much change)
We have
We have
9
We have
So we have
Error correction:
10
Activation
where link /part
11
For hidden neuron:
Sequentially going backwards tracking the error signals from
all the neurons to the hidden neuron in question.
12
𝜕𝜕𝜀𝜀(𝑛𝑛)
𝜕𝜕𝑒𝑒𝑘𝑘 (𝑛𝑛)
13
We have
−𝑒𝑒𝑗𝑗 (𝑛𝑛)
14
Backward propagation of error:
15
Correction (delta rule):
Forward pass:
or
Hidden Output
For every input
Backward pass: pattern / vector
Error / Local gradient flow from output
layer towards input via the hidden layers
16
Error Functions
By considering
Invoking log-
likelihood
17
*discuss later
18
Needs to be continuous
Logistic function:
19
Output neuron
Hidden neuron
Hyperbolic tangent:
20
Output neuron
Hidden neuron
21
Oscillatory, unstable network!
Start time
Connection Dependent*
23
No analytical convergence proof
One way:
24