It's possible to hack it but I would not recommend it.

If your concern is higher-resolution checkpoints than a generation  
boundary, which suggests very long evaluation times, you may wish to  
go with Steady-State evolution instead.  By default  
SteadyStateEvolutionState checkpoints every N*M evaluations where N  
is the size of the initial population.  But you can modify the  
generation boundary definition in the code more easily than in the  
generational evolution procedure.  A hack like changing:

             // COMPUTE GENERATION BOUNDARY
             generationBoundary = (evaluations % generationSize == 0);

to

             // COMPUTE GENERATION BOUNDARY
             generationBoundary = (evaluations % 50 == 0);  // do  
every 50 evals

... might do the trick in Steady State.

Sean

On Jun 2, 2008, at 12:02 PM, Shane wrote:

> Is it possible to save a checkpoint in the middle of a generation?