[mlpack] GSOC 2021 project ideas

[Ryan Curtin]

Hi Anmolpreet,

These sound like nice ideas.  It seems likely to me that it would be
possible to implement more than one of these in the summer, but also you
should be sure to think about example usages: if we have many different
quality metrics, that is great, but what is maybe more important to
drive users to them is coherent and clear examples demonstrating their
usage.  So you may want to ensure that documentation and examples are
part of the plan also. :)

I hope this is helpful!  I'm not deeply involved with #2294, so others
may have other comments too.

Thanks!

Ryan

On Sat, Mar 13, 2021 at 11:55:22PM +0530, Anmolpreet Singh wrote:
>    Hello all!
> 
>     
> 
>    Being with MLpack community for some time I realized a few ideas which can
>    be implemented during period of GSOC 2021. Taking into consideration the
>    shorter time this year I want to propose the following idea of
>    implementing  some useful metrics of ML as a GSOC 2021 summer project.
> 
>     
> 
>    The structural similarity index measure (SSIM) is a method for predicting
>    the perceived quality of images or measuring the similarity between 2
>    images considering the structural information. It has good applications in
>    image compression (checking quality of compressed image), image
>    restoration and pattern recognition. I think it will be some interesting
>    stuff to add in MLpack, as it is an important part of image processing.
>    Also, its need is highlighted from the issue #2294(Addition of essential
>    metrics only) which is pending due to issues in implementation. So, it
>    will be good if is done in an organized way under guidance of mentors. I
>    have also gone through a couple of recent research papers regarding
>    improvements in this.
> 
>     
> 
>    PSNR is another image quality metric which has wide use in digital image
>    processing. I feel that Metrics like these which have tremendous use will
>    fit good in MLpack. Also, discussion will bring more metrics into
>    consideration.
> 
>     
> 
>    In addition, I may also add metrics like Top K accuracy and some other
>    needful metrics (suggestions are welcomed) which may help in evaluating
>    the performance of various models. Recently,  I have done a data science
>    project (chat-bot using NLP)  and planning to continue my journey with
>    this idea.
> 
>     
> 
>    Any feedback or suggestion for these ideas will be really helpful for
>    further planning this according to the discussion.
> 
>    Regards,
> 
>    Anmolpreet Singh
> 
>     

> _______________________________________________
> mlpack mailing list
> mlpack at lists.mlpack.org
> http://knife.lugatgt.org/cgi-bin/mailman/listinfo/mlpack


-- 
Ryan Curtin    | "We need some time for some things to happen!"
ryan at ratml.org |   - Bells