GMDH-type neural networks

6
GMDH-type neural networks 1

description

GMDH-type neural networks. GMDH. Inductive Learning Algorithms for Complex Systems Modeling Alexey G. Ivakhnenko (1994) Group Method of Data Handling Inductive learning “Survival of the fittest” Fully automatic structural “Self-organization” “external criterion” - PowerPoint PPT Presentation

Transcript of GMDH-type neural networks

Page 1: GMDH-type neural networks

GMDH-type neural networks

1

Page 2: GMDH-type neural networks

GMDH• Inductive Learning Algorithms for Complex Systems Modeling • Alexey G. Ivakhnenko (1994)

• Group Method of Data Handling• Inductive learning

• “Survival of the fittest”• Fully automatic structural

• “Self-organization”• “external criterion”

• Parametric optimization

2

Page 3: GMDH-type neural networks

Unit neuron• Polynomial neural networks• Solving the nonlinear problem using polynomial instead of linear

combination.• Kolmogorov-Gabor polynomial

3�̂�

𝑥𝑖

𝑥 𝑗

𝑓 (𝑥𝑖 ,𝑥 𝑗 )

Page 4: GMDH-type neural networks

• Weight update:

Back propagation:1-hidden-layer:f(XW)2-hidden-layer:f(f(XU)W)

4

U W

Error ε

∆ u=−α 𝜕 ε𝜕u=−𝛼∑ 𝜕 (𝑦− 𝑓 ( 𝑓 ( 𝑋𝑈 )𝑊 ) )𝜕 𝑓 ( 𝑓 ( 𝑋𝑈 )𝑊 )

𝜕 𝑓 ( 𝑓 ( 𝑋𝑈 )𝑊 )𝜕 𝑓 ( 𝑋𝑈 )

𝜕 𝑓 ( 𝑋𝑈 )𝜕𝑢 =𝛼∑ ( 𝑦− 𝑓 ( 𝑓 (𝑋𝑈 )𝑊 )) 𝜕 𝑓 ( 𝑓 (𝑋𝑈 )𝑊 )

𝜕 𝑓 (𝑋𝑈 )𝑥

Page 5: GMDH-type neural networks

External Criterion• Principle of self-organization

• Regularity criterion• It consists of a squared error from testing set B

• Symmetric Regularity criterion5

complexity

delta

Δ(B)

ε(A+B)

A: training setB: testing setΔ(): external criterion

Global minimum

Page 6: GMDH-type neural networks

Initial model selection

6External criterion

Optimal estimation