Better descriptions for random initializations

[mstimberg]

With the polishing branch, intializations that refer to rand() or randn() already look a bit nicer. But I think we could use this as a first example of automatic annotation with additional analysis.
If an expression only contains a single rand() call, we can use sympy's substitution mechanism and substitute it once with 0 and once with 1, and then use this to display the lower and upper bounds. So if you have an initialization like:

v = El + (V_th - El)*rand()

This would currently be displayed as:

but we could further annotate this with something like

We could probably do something fancy with sympy as well to get the mean and variance of a normal distribution if the expression uses randn().

[Lavina-korani]

A possible solution would be to automatically analyze initialization expressions that contain random functions like rand() or randn(). Using symbolic analysis with SymPy, the system could interpret the expression and provide additional annotations to help users understand the distribution of values.

For expressions with a single rand() call, the tool could substitute rand() with 0 and 1 to compute the minimum and maximum values, allowing it to display the range of the initialization. For example, an expression like v = El + (V_th - El)*rand() could be annotated to show that the resulting values lie in the range [El, V_th].

Similarly, if the expression contains randn(), it could analyze the expression to estimate statistical properties such as the mean and variance of the resulting normal distribution. This would make the initialization clearer and more informative for users inspecting the model. I can elaborate it more... and possibly solve this issue.

[Utkarsha2811]

Hello,

I’m interested in working on this issue and improving the descriptions for random initializations.

From the description, it seems related to how expressions using rand() or randn() are represented when exporting models. I explored the repository and noticed the baseexport module, which looks related to exporting functionality.

Could you please confirm if this is the correct place to implement the improvement, or suggest where the change should be made?

Thank you!

[Utkarsha2811]

PR for the issue is Under Review: #63

[realanu0812]

Hi @mstimberg,

This sounds like a great direction. I’d like to work on adding a small SymPy-based annotation step for expressions with a single rand() call, using substitution (rand() → 0/1) to infer lower and upper bounds.

I can start with a minimal implementation for rand() and extend it to randn() (mean/variance) if that looks useful.

I’ll put together a small prototype and open a PR soon.

Best,
Anurag

[realanu0812]

Hi @mstimberg,

I’ve opened a PR implementing a first version of the SymPy-based annotation for initializer expressions containing a single rand() call.

It infers lower and upper bounds via substitution (rand() → 0/1) and displays them as a Uniform range in the markdown export. I also fixed an issue with array-based indices in expand_initializer that was causing errors during rendering.

I’ve kept the scope intentionally minimal for now. I’d appreciate your feedback on whether this direction looks good and how you’d like to extend it further.

Thanks,
Anurag