Hi -
I find there is a ‘hole’ in documentation between older and newer versions of Nengo. e.g. probes are used all the time in earlier versions, but I suppose the ability to right click on ensemble, states, etc. has made much of that unnecessary.
I want to record data from my models that I can’t figure out how to get data out of the gui to analyze. A trivial example would be spike data (I can download values and voltages into .cvs files, but not spike data)
I can create probe objects. But all of the example of using the data from the probes involve running simulations that one can’t do with the gui. haven’t been able to access data stored in the probe while using the gui. Could you lead me to an example (e.g. a msec by msec output of values from a semantic cloud)
Thanks
Howard
Hi Howard,
You’re right; at the moment, the GUI doesn’t do anything with probes. When you make plots by right clicking on objects and choosing a plot type, the GUI sets up some internal machinery to do something that is similar to what probes do, but not exactly the same. Because of that, not all of the plots are set up to export their data to CSV. I’ve made a new issue on the GUI bug tracker so that we remember to fix that in the future: https://github.com/nengo/nengo_gui/issues/874
For now, anything that is shown in the GUI can be done outside of the GUI using probes and analyzing the probed data after the fact. The semantic cloud plot, for example, highlights semantic pointer labels according to their similarity to the known pointers in the vocabulary. That similarity can be computed from probed data also; a good example to see how to do this is the associative memory example available at http://pythonhosted.org/nengo/examples/associative_memory.html
Hope that helps,
Trevor