Hi John,
It seems really a nice Idea to have machine Learning integrated with
ASCEND. I am very excited to contribute and will definitely plan for that
,but I believe that first step for that would be to implement and Develop
support for different optimization algorithm and then integrate these
into Machine Learning in latter stage .
I have few question regarding Optimization solver
Does ascend have model for big and complex matrix manipulation like
"Inverse" ? If we plan for Implementing some complex Optimization
Algorithm than It would be essential. Actually I was Just playing with
ascend code by Implementing few Machine learning algorithm on it, but I
was unable to locate any such model.
Why are you planning to implement PSO and Newton gradient ascent in C ?
It would be lot easier to implement same in Python by using NumPy and SciPy
library .
On Mon, Mar 16, 2015 at 5:17 PM, John Pye <[email protected]> wrote:
> Hi Kumar
>
> Thanks for your interest in our project. It would be very interesting to
> have your thoughts on bringing aspects of machine learning to the ASCEND
> modelling environment. My feeling would be that it's quite important to
> first understand the type of non-linear modelling that ASCEND immediately
> seeks to support, and then to consider ways of building on top of that.
>
> Perhaps an area that might interest you would be in parameter estimation
> from noisy data? For example, if someone has a correlation with a certain
> number of unknown parameters (eg y = mx+b) as well as a data set for (x,y),
> can you implement a general-purpose solver in ASCEND that would give
> estimates for (m,b) together with their statistical uncertainty based on
> the data? If a number of different possible correlations are available (eg
> y = ax^2 + bx + c) can you provide an assessment of which one has the best
> performance at the same time as not having so many parameters as to be
> 'over fitting' the data? This is not a new problem; there are well
> developed techniques that would apply, but we don't currently have any
> support for this mode of calculation in ASCEND. Perhaps you could develop
> the basic approach, and then implement some machine learning stuff on top
> of that?
>
> Cheers
> JP
>
>
> On 15/03/15 06:02, Kumar Shubham wrote:
>
> Hi Guys,
>
> I have previously worked on Particle swarm optimization and other Bio -
> Inspired algorithm for cross validation using newton's gradient based
> optimization technique in one of my paper on optimizing the performance of
> Twin support Vector Machine, which recently has been selected for
> publication in Journal of Machine Learning and Cybernetics. Although, I
> have done that in Python and Matlab, but doing it in Ascend platform will
> be new for me.
>
> I am really excited to contribute to Ascend, either in GSOC or after
> that!!.
> I have already gone through example, can you please guide me through
> next step as I am more inclined to optimization based algorithms.
>
> --
> =====================================================
> ~"Go over, go under, go around, or go through. But Never give up "~
>
> Regards,
> *Kumar Shubham*
> Pre final Year Student
> B.Tech: ELECTRONICS AND COMMUNICATION
> The LNM Institute of Information Technology, Jaipur, India
> Alternate Email ID: [email protected]
> WebSite : http://kumarshubham.wix.com/project
> Linkedin: https://www.linkedin.com/profile/view?id=248505557
>
>
>
--
=====================================================
~"Go over, go under, go around, or go through. But Never give up "~
Regards,
*Kumar Shubham*
Pre final Year Student
B.Tech: ELECTRONICS AND COMMUNICATION
The LNM Institute of Information Technology, Jaipur, India
Alternate Email ID: [email protected]
WebSite : http://kumarshubham.wix.com/project
Linkedin: https://www.linkedin.com/profile/view?id=248505557
====================================================
------------------------------------------------------------------------------
Dive into the World of Parallel Programming The Go Parallel Website, sponsored
by Intel and developed in partnership with Slashdot Media, is your hub for all
things parallel software development, from weekly thought leadership blogs to
news, videos, case studies, tutorials and more. Take a look and join the
conversation now. http://goparallel.sourceforge.net/
_______________________________________________
Ascend-sim-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ascend-sim-users
