Functions for multi-dimensional data

[drahnoel]

When performing parameter sweeps, the data gets multidimensional. For post-processing/visualization it seems that formulas for handling this are missing, forcing people to use octave code.

I propose two features, which are independent of each other:

• Proper indexing: eg.: variable[index,1:5,5:,:5]
  A number or integer variable eliminates the dependency. A range is applied to the data and to the dependent variable. Therefore, we have also a need for:
  • get_dependent(variable, “name”)
    Get the dependent of data with given name. Is usually equivalent to using name directly, unless a range was applied.
• apply_on_dependent(variable, dependent_name, function)
  Return data with the dependent_name dependency eliminated. It is eliminated by calling function on each 1-D data with dependent_name as only dependency. (function could be a string to be evaluated.)
  This would enable all kind of usefull operations. Ideas for function:
  • min, max, average, median, quantile
    Very interesting eg for Monte Carlo or parameter sweeps.
  • Just indexing: eg: data[3]
    For example: Simulation is Harmonic Balance + Parameter Sweep. Using this, we can do apply_on_dependant(variable, “frequency”, “data[3]”) to get the 2nd harmonic over the parameter sweep.
  • interpolate(data, value, kind=”linear”)
    Idea is to remove the dependency with dependent_name by interpolating the data accross the dependent variable.

[Carsten]

Hi, drahnoel,

> A range is applied to the data and to the dependent variable…..
you probably mean _in_dependent variable, right?

>get_dependent(variable, “name”)
>Get the dependent of data with given name. Is usually equivalent to using name directly, unless a range was applied.

I don’t understand this. Could you elaborate please?

Cheers
Carsten

[drahnoel]

Yes you are right, I will edit.

Let’s say we have variable “signal” depending on variable “frequency”.

With the proposed syntax, you could do something like:

(signal/frequency)[2:3]

or

signal[2:3]/frequency[2:3]

What I propose is:

s = signal[2:3]

s/get_dependencies(s, “frequency”)

Now, that I think about it, this does not seem too usefull, as long as the dependencies are always bound to variables.

[drahnoel]

I cannot edit anymore

[Author]

Posts