GSOC Idea and Query

[Kuldeep Borkar]

I was planning to be a part of GSOC this summer with SymPy to learn more and
implement something big.

For this,
I was planning to work on improving SymPy's stats module;
I checked there are many good things already implemented in the stats module.
And, I found issue(https://github.com/sympy/sympy/issues/17197)  about the Random Walk implementation which is in progress, but it seems the issue created is closed for now and I am looking forward to complete it.

On the ideas page of SymPy for GSOC in the Probability section I found things which can be implemented as a part of GSOC project.
I saw things there which interests me to work on:

--> Reproducibility of Sampling Outputs of Stochastic Processes,
WienerProcess(I think the idea of ito calculus is covered here),
Completing Random Walk(I think it was a typo there it should be Prototype not Protorype), etc.

But my query is:
I was planning to implement something different;

A better way to use probability of events like one mentioned here(https://github.com/sympy/sympy/issues/20111):
--> Working with events rather than random variables.
--> Currently I don't think there's is a way to define just an event and
not specifically a random variable like
A, B = event('A, B')
P(A) = x
P(B & A) = y
P(B & !A)?

Also, currently there is not a way to assign probability like a certain event A

has probability P(A) = x
(Please enlighten me if I am making a mistake somewhere ; )  )
and noise processes too, like white noise (White noise refers to a statistical model for signals and signal sources, rather than to any specific signal), 

is it possible to work on some of the things from ideas page and some from our own?

[I am trying to understand the codebase now(Are there any tips which could help me to understand things more efficiently? )  : (
It may not be a valid question though, but thought I should try to ask first.]

[Kuldeep Borkar]

Hello SymPy Community,

Few days ago, I was having a discussion regarding my GSOC Project Idea with this year's mentor for the Probability Section and on my idea regarding implementation of noise processes I came up with the suggestion of adding a method to Visualize the noise processes, also the Random Walk which I am planning to complete as a part of my GSOC this summer.

My suggestion to visualize was to add a method like .visualize( ) to Noise Processes and Random Walk and pass some flags whether the user wants it in animated sort of thing or not but this idea was 

Request For Comments(RFC) from the SymPy community since .visualize( ) might not be the best the to implement this thing.

So any feedback/comments regarding the above would be helpful a lot : )

[Peter Stahlecker]

I have never used sympy statistics, hence my comments may be without any use.

1. 

With matplotlib there is an excellent visualisation library available.

Hard for me to see, how you can beat it.

2.

There is a library sdeint available, which numerically integrates Ito or Stratchonovich stochastic differential equations.

Hence, it gives sample paths of integrated white noise for free, of course.

(I think, it is a bit of an one man show, not really being developed much, but the stochastic integration works)

 

--
You received this message because you are subscribed to the Google Groups "sympy" group.
To unsubscribe from this group and stop receiving emails from it, send an email to sympy+un...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/sympy/82761518-db7a-4f23-9e72-5549c5986fabn%40googlegroups.com.

[Kuldeep Borkar]

Thank you for the response Sir,

1. It's just the opposite, I am not even trying to beat matplotlib, I am just trying to develop what's already there in SymPy .

In SymPy presently the plots are rendered using matplotlib as a backend.

If Random Walk is being implemented in SymPy user would like to visualize it in the most easy and efficient way possible.

It's like either use matplotlib or in SymPy just use something like RandomWalk('rw', animated=True, store='C:\\').visualize. So, what would be better?

2. Thank you for telling me about sdeint library.

I think since, stochastic processes is already implemented in SymPy so we should make it a full-fledged stat's module.

For the point that we can get the sample paths of integrated white noise for free so yes it might not be the novel idea to implement that in SymPy so for that we can add more functionality like, from the given sample's identify does that belong to a noise process or not(return True if it belongs and False if it does not), I don't think this is already implemented.

This way we could extend the stat's module and don't just implement what's already implemented.

[Peter Stahlecker]

As I said, I have never uded this stats module, hence no idea what is does.

There is a module lmfit, which I have played around with a very little bit.

It is about fitting noisy data to some 'curve'.

No idea if this is similar to what you want to do.

To view this discussion on the web visit https://groups.google.com/d/msgid/sympy/2d94a0de-d6a1-4a40-9855-ca1cb8ee056dn%40googlegroups.com.

--

Best regards,

Peter Stahlecker

[Kuldeep Borkar]

Thank you Sir Alan for telling me about the Asymptote software, it surely is better than matplotlib(it would have been best if the scripting language was python), I will see if I could use that in visualizing Random Walk and Noise Processes : )

No Sir @Peter, this is different from what I am trying to implement.(Thanks for letting me know about different libraries though.)