Learning’98, UC3M, Madrid, Spain

Index

Networks to estimate ‘a posteriori’ class probabilities

Previous approaches

Example

Fundamentals

1.- The binary one-dimensional Logistic Problem (OPDE-LOP)

1.- The binary one-dimensional Logistic Problem (OPDE-LOP)

2.- A multidimensional case.The Softmax Perceptron (OPDE-SP)

3.- A general multidimensional. case.The Generalized Softmax Perceptron (OPDE-GSP)

Particular case of Central Distribution (OPDE-GSP)

Gradient based learning rule (OPDE-GSP)

Gradient based learning rule (OPDE-GSP)

Results

DE vs PE: convergence speed

DE vs PE: generalization

Conclusions

Further lines