Gsoc 2012 Idea

[sachin004]

Hi,

Introduction: I am currently a third year computer science undergraduate from Bits-Pilani , and I would like to participate in sympy development in GSCOC 2012.

Experience: I have been programming in python since 2-3 years and I had a very good experience with it, I have been using numpy, scipy, wxpython since many days, and sympy since past 3 months and I have been working with the source code of sympy since the past 15 days .I have been using sympy for a project on “advanced numerical methods” since past three months.

Project Idea:

The following are the project ideas that I would like to work on,

1. Implementing quantifiers:

  i.Universal quantification

  ii.Existential quantification

 iii. Multiple ordered quantifiers

2. Binary Predicates

 3. Linear regression

any suggestions or improvements are greatly welcomed

Regards

Sachin Irukula

[sachin004]

Hi everyone,
   Any suggestions

[Matthew Rocklin]

Hi Sachin, 

I'm not knowledgable about SymPy's logic system although there has been some discussion of this topic on this listhost recently. I would perform a search on the mailing list to find the recent e-mail conversations. I think that there is  some work to do here but I don't know any more.

I am more interested in your third idea but I'm not sure how it connects to SymPy. How do you think SymPy can be useful for linear regression? Usually linear regression is done from a data point of view, not a symbolic one. Maybe this is more appropriate for R/SciPy/Pandas? I don't know. Can you explain your thoughts on this idea?

-Matt

Hi everyone,
   Any suggestions

--
You received this message because you are subscribed to the Google Groups "sympy" group.

To view this discussion on the web visit https://groups.google.com/d/msg/sympy/-/M0spM2cJ3sYJ.

To post to this group, send email to sy...@googlegroups.com.
To unsubscribe from this group, send email to sympy+un...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/sympy?hl=en.

[sachin004]

Hii matthew,

Firstly thanks for the suggestions.
I have suggested linear regression just as a stepping stone to symbolic regression. Even though both are different in many ways what i would like to suggest is that sympy to support symbolic regression (which I thought of including based on the response from the mentors or others)and along with it to provide support for linear as well as non linear regression.I have mentioned only few things to know more from you. If you like I would i would give detailed explanation on symbolic regression.

Regards
Sachin Irukula

[Aaron Meurer]

Yes, please do. I'm not clear what exactly you mean by symbolic regression.

Aaron Meurer

>
>
> Regards
> Sachin Irukula
>
> --
> You received this message because you are subscribed to the Google Groups
> "sympy" group.
> To view this discussion on the web visit

> https://groups.google.com/d/msg/sympy/-/cN4YxFsYGyUJ.

[sachin004]

symbolic regression (or symbolic function identification)can be done by genetic programming (many other methods are available ). symbolic regression finds the symbolic expression function to the given data input and outputs and outputs an expression best fitted for the inputs. the basic difference between symbolic regression and normal regression is normal regression assumes a model(expression)  and determines the coefficients, where as symbolic regression searches for the model and fits it.
    Currently mathematica, matlab and many more are supporting(implemented) symbolic regression.

http://library.wolfram.com/infocenter/Conferences/5392/
http://sites.google.com/site/gptips4matlab/

[Ronan Lamy]

That's interesting. However, sympy doesn't deal with data, so I think
that this should be an external project depending on sympy and pandas
(http://pandas.pydata.org/ ) and/or scikit-learn
(http://scikit-learn.org/stable/ ).

[sachin004]

does it mean symbolic regression doesn't come under a project for sympy gsoc

Regards
Sachin

[Ronan Lamy]

Le mardi 27 mars 2012 à 10:57 -0700, sachin004 a écrit :
> does it mean symbolic regression doesn't come under a project for
> sympy gsoc

I don't know. It doesn't seem farther from sympy's core goals than the
Live/Gamma or Android projects. The difficulty is in knowing the
dependency requirements up-front and finding how it can fit with
existing projects. Another issue is to find a suitable mentor.

[sachin004]

ok, but how can i know whether there is any suitable mentor for this topic or the other topics that i have mentioned in my previous post ? As I have less time and with current mid term examinations for me its bit difficult if i don't finalize on my project proposal.
Regards
sachin

[sachin004]

and i would like to add

implementing de Moivre's formula to my ideas list.

[sachin004]

sorry for that I didn't realize that it's already been implemented

[sachin004]

4.Refactoring old handlers in assumptions.
          Partial work has already been started.

[sachin004]

Implementation of universal quantifiers

Universal Quantification:  Function: for_all(expr,variables,condition)

·       ∀ _xexpr which says that expr holds for all values of x for this the function looks like for_all(expr,var) where var=Tuple(x)

this returns an assertion based on the expr and the values of x.

·        ∀_{x1,x2,...}expr which says that expr holds for all values of x for this the function looks like for_all(expr,var) where var=(x1,x2,x3….) this returns an assertion based on the expr and the values of x1,x2,x3...

·        ∀_x,condexpr which says that expr holds for all values of x which satisfy the condition for this the function looks like for_all(expr,var) where var=(x1,x2,x3….) this returns an assertion based on the expr and the values of x1,x2,x3...

[Aaron Meurer]

I think you always need the condition. Just saying "for all x"
doesn't make any sense. You have to have "for all x in some set".

By the way, it's just a semantics things, but expr should really be
called cond, since it needs to be a boolean condition, not just some
generic expression.

Aaron Meurer

> --
> You received this message because you are subscribed to the Google Groups
> "sympy" group.
> To view this discussion on the web visit

> https://groups.google.com/d/msg/sympy/-/X7lB_Ir0NLcJ.

[sachin004]

Ok, If we are calling expr as cond then what shall we call condition as.
Regards
Sachin

[sachin004]

Any more suggestions/topics that i can add to this idea.

Regards
Sachin

[sachin004]

[Ronan Lamy]

Le mercredi 28 mars 2012 à 22:05 -0700, sachin004 a écrit :
> Ok, If we are calling expr as cond then what shall we call condition
> as.

If you have two parameters with the same name, it's a sign that there's
a problem with your design. Either the parameters should be combined or
their meaning should be clarified.

[sachin004]

Hi everyone,

  I would like to know which algorithm would be better for checking the satisfiability of first order logic expressions, I went through simplify theorem prover which seems to be good, and also is there any potential mentor for this area(logic module).

[sachin004]

Gsoc Proposal : https://github.com/sympy/sympy/wiki/GSoC-2012-Application:Sachin-Irukula-:-Implementation-of-Quantifiers-and-Cylindrical-algebraic-decomposition-algorithm

Regards

Sachin Irukula