Simulated Annealing for Optimality Theory
Introduction
Run a
simulation
Examples
Dutch stress
Dutch stress
(CLIN 2004)
Literature
Author:
Tamás Bíró
SA-OT: examples

Examples

Gush! This looks interesting! But how should I fill in this form?

Click here to try out the model for Dutch stress.


This is an example where SA-OT works. Try out very different t-step values. With higher ones, you will be stuck quite often in the local optimum C. Whereas lowering the t-step values will increase the precision, that is, the algorithm will return the global optimum A with a higher probability.


Unranked tableau:

  & Con1 & Con2 & Con3
A &      &      &  1
B &   2  &      &
C &   1  &      &


Neighbourhood structure:

A: B
B: A & C
C: B


K-values:

Con1: 2
Con2:   1
Con3 : 0


K-max:  3  
K-min:   -2 
K-step:   1 
t-max    3
t-min:   0 
t-step:  0.5  

Number of runs: 50


The following tableau looks very similar, but you will always receive a 50%-50% distribution between the two local optima. For an explanation, read my dissertation.


Unranked tableau:

  & Con1 & Con2 & Con3
A &      &      &  1
B &   2  &      &
C &      &  1   &


Neighbourhood structure:

A: B
B: A & C
C: B


K-values:

Con1: 2
Con2:   1
Con3 : 0


K-max:  3  
K-min:   -2 
K-step:   1 
t-max    3
t-min:   0 
t-step:  0.5  

Number of runs: 50