Questions concerning values and crashes in Octopus

Hello world.

I have some issues for which I would really enjoy some help...

I don't really find much information about such issues on the internet.

1. Why is my Number-Input postive and the graph in Octopus negativ and Vice-versa (see Picture)

Is this the way it is supposed to be? Is the programm seeking for the minimum Number-Values or minimum in the graph?

2. How does 'Troubleshooting' work? (Is there an extended manual or so?) I need it as my system crashes, and I assume this is for certain individuals...

2b. Does Octopus have any Problems dealing with 0? Or a change from positive to negative numbers during the process?

3. What should the populationsize depend on?

If it takes rather long to calculate one individual, should I lower the number of individuals?

Thank you very much

Replies to This Discussion

Permalink Reply by Robert Vier on August 16, 2014 at 2:15

hi andreas.

1) the number in the graph shows the bounds of the current set of solutions displayed in octopus. if you reinstate a solution from the graph, its fitness values should lie within the domains on the axes. it might be that the solution you are looking at is one that octopus tried out, but has not made it into the front- or elite-set - hence it is not displayed and the fitness value is not used for scaling the axe's domains.

2.1) in case your system crashes during the search, you can check the box 'save gh file before evaluation of each individual to track sudden crashes of rhino caused by some special cases in the definition', which does the following:

- generate a new set of parameters

- set the parameters into the number sliders / gene lists

- save the gh-file [this is what is turned on by the checkbox, which comes at a performance cost of course]

- expire the number sliders / gene lists

- recompute the required parts of the definition

.. which, after a crash, should give you the possibility to assess the cause of the crash yourself.

2.2) troubleshooting saves the individuals which exceed the max.evaluation time set on the navigation page. you can then check why they take so long

2b) no, there should not be any problems since the domains of objective values always are being normalized to ]0;1[

3) this is tricky, and depends mostly on the complexity of the problem. if there are many parameters, the pop-size should be higher to cover a larger portion of the design space just by the initial random tries of generation 0. also, the chance of a more diverse set of solutions is higher with larger population size.

best

Robert

RSS