systemml-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Madison Myers <madisonjmy...@gmail.com>
Subject Re: Documentation Personas
Date Fri, 30 Sep 2016 16:05:57 GMT
*Although these types of meetings exclude the community, I believe
theymight still bring value, but I would recommend two things :- Get all
the participants to subscribe to the mailing list- After the meeting,
please provide a summary with the main ideas andfinish the discussion
around agreement on the next steps here on themailing list (Apache has a
say that, if it didn't happen on the mailinglist, it didn't happen.)*

I am sorry that you feel that this meeting excludes the community. I
actually am surprised as I have been a big advocate of communication since
the beginning! I actually dm'd you the other day asking to be looped in as
several things have been changed on the website without notifying the team
who is designated to work on it.

Moreover, not sure if you know, but the design team actually interviewed
several people from the SystemML team in an effort to better understand
their work flow, needs, etc. and that is what we are coming together to
discuss. Therefore I don't think it excludes anyone at all.

I've been trying to get on the dev list for some time, but because there
was some confusion, I was not able to communicate on here until this time.
>From here on out, as I said in the other thread, I will be updating
everyone on everything.

With all that being said, I am happy you value communication as much as me
and we can work together to make sure everyone is on the same page and has
a voice.









*If a helloworld is not basic enough, maybe we need to identify what
iswrong with the usability of SystemML. Let's give an example of a
productthat I worked in the past, we were building a getting started guide
usingeclipse, and we identified that there were multiple steps that were
varyhard and required folks to have a deep technical understanding of
theinternal of the product to be able to add the required dependencies
andother steps, with the problem identified, we built an eclipse plugin
thatsolved much of the issues. See the guide as an example here
:http://tuscany.apache.org/getting-started-with-tuscany-using-tuscany-eclipse-plugin.html
<http://tuscany.apache.org/getting-started-with-tuscany-using-tuscany-eclipse-plugin.html>*

I think both. You are completely right. SystemML does need to be worked on
to be more user friendly! I also think a very basic "this is how you
literally download systemml and start it" is needed in addition to other
tutorials such as the Jupyter notebooks (they can walk people through
uploading data, matrices, algorithms, etc.).






*+1, Notebooks seems to be a simple way to interact with code and a UI
thata lot of data scientists would be familiar with.If we have a detailed
guide (a.k.a. Tutorial) that explained how to addsystemml dependency, write
a matrix multiplication, and execute... thiscould be a "helloworld" that
people could easily understand, no ?*

See above! I totally get where you're coming from, but I encourage us to
have both. Coming from the perspective of a new data scientist (and having
asked my peers to try SystemML) the overwhelming feedback is that people
don't know how to use it and that they would like more tutorials that are
clear and easy to follow.

Thanks so much for taking the time to give your input! Really appreciate
it. Will update you after the meeting.

On Thu, Sep 29, 2016 at 11:12 PM, Luciano Resende <luckbr1975@gmail.com>
wrote:

> On Thu, Sep 29, 2016 at 8:10 PM, Madison Myers <madisonjmyers@gmail.com>
> wrote:
>
> > Thanks for your input!
> >
> > I have an all day meeting/brainstorm session with the design team
> tomorrow
> > so this is super helpful. I totally agree with Mike that the personas
> > should not be directly addressed, but rather opportunities for different
> > personas/levels to interact with SystemML are available on the website
> for
> > the different user personas to navigate.
> >
> >
> Although these types of meetings exclude the community, I believe they
> might still bring value, but I would recommend two things :
> - Get all the participants to subscribe to the mailing list
> - After the meeting, please provide a summary with the main ideas and
> finish the discussion around agreement on the next steps here on the
> mailing list (Apache has a say that, if it didn't happen on the mailing
> list, it didn't happen.)
>
> I think starting out, we are going to address the main issue: converting
> > new users to SystemML. Unfortunately even just a how to write your
> > algorithms is not a basic enough "hello world" if we want data scientists
> > to start using SystemML- especially those in school. Because more
> advanced
> > users can navigate the current documentation a little easier, we will
> focus
> > on that later, though it remains a priority.
> >
> >
> If a helloworld is not basic enough, maybe we need to identify what is
> wrong with the usability of SystemML. Let's give an example of a product
> that I worked in the past, we were building a getting started guide using
> eclipse, and we identified that there were multiple steps that were vary
> hard and required folks to have a deep technical understanding of the
> internal of the product to be able to add the required dependencies and
> other steps, with the problem identified, we built an eclipse plugin that
> solved much of the issues. See the guide as an example here :
> http://tuscany.apache.org/getting-started-with-tuscany-
> using-tuscany-eclipse-plugin.html
> .
>
>
>
> > I can give you an update after the meeting tomorrow, but for now, we
> > generally spoke about making tutorials and documentation clearer and even
> > potentially having links to Jupyter notebooks where users could follow
> > along. We will also be focusing on the website flow so that people don't
> > get lost while navigating and so that selections are more intuitive.
> >
> >
> +1, Notebooks seems to be a simple way to interact with code and a UI that
> a lot of data scientists would be familiar with.
>
> If we have a detailed guide (a.k.a. Tutorial) that explained how to add
> systemml dependency, write a matrix multiplication, and execute... this
> could be a "helloworld" that people could easily understand, no ?
>
>
> > If you have any tutorials that are not already on the website, please
> send
> > them my way! We want to compile everything in order to have it all in one
> > place (and more organized!).
> >
>
> +1
>
>
> >
> > Thanks!
> >
> > On Thu, Sep 29, 2016 at 5:51 PM, <dusenberrymw@gmail.com> wrote:
> >
> > > These topics and the idea of the personas are great!  I would suggest
> > that
> > > we don't explicitly use the specific persona titles though. I.e.
> Instead
> > of
> > > specifically using "data scientists" we could just use "machine
> learning"
> > > such as "Getting started guide to machine learning with SystemML".
> > Likewise
> > > the topics for "data engineers" could instead be something like
> "Getting
> > > started guide to engine development". We could also have a "deployment"
> > > guide, etc.  In each we could also have beginner and expert sections.
> > >
> > > --
> > >
> > > Mike Dusenberry
> > > GitHub: github.com/dusenberrymw
> > > LinkedIn: linkedin.com/in/mikedusenberry
> > >
> > > Sent from my iPhone.
> > >
> > >
> > > > On Sep 29, 2016, at 4:07 PM, Luciano Resende <luckbr1975@gmail.com>
> > > wrote:
> > > >
> > > > On Wed, Sep 28, 2016 at 6:06 PM, Madison Myers <
> > madisonjmyers@gmail.com>
> > > > wrote:
> > > >
> > > >> Thanks for opening up this dialogue Felix & Luciano!
> > > >> .....
> > > >>
> > > >> Additionally, Luciano mentioned that he was trying to find two key
> > > personas
> > > >> for the SystemML website. Looking at the results from the SystemML
> > > survey I
> > > >> sent out a few weeks ago, it is clear that an overwhelming majority
> of
> > > >> people voted for Data Scientist- new and Data Scientist- advanced.
> > > Input on
> > > >> if you agree on these results would be helpful.
> > > >>
> > > >> ....
> > > >>
> > > >> Thanks again!
> > > >> Madison
> > > >
> > > >
> > > >
> > > > I was thinking on two main personas : Data Scientist and Data
> > Engineers.
> > > > And at least for the Data Scientist, we might have to have topics for
> > > both
> > > > R and Python.
> > > >
> > > >
> > > > In summary, I would say :
> > > >
> > > >
> > > >
> > > > Getting Started for Data Scientists (note that we should have a
> version
> > > for
> > > > R and Python):
> > > >
> > > > - Creating algorithms with SystemML : this is kind a helloworld
> where a
> > > > Data Scientist creates and runs a very simple algorithm.
> > > >
> > > > - Running existing algorithms with SystemML : in this scenario, we
> > would
> > > > guide the data scientist on how to run an algorithm from the sample
> > > library
> > > > from SystemML. We should also guide the data scientist to choose the
> > > right
> > > > algorithm for his/her specific problem.
> > > >
> > > > - Customizing existing algorithms with SystemML: in this scenario, we
> > > would
> > > > guide the data scientist to start making customization and applying
> the
> > > > algorithm for different data sets.
> > > >
> > > >
> > > >
> > > >
> > > > Getting started for Data Engineers:
> > > >
> > > > - Overview of different options/runtimes supported by SystemML
> > > >
> > > > - Debugging SystemML ???
> > > >
> > > > - Engine development guide ???
> > > >
> > > > Thoughts ?
> > > >
> > > >
> > > > --
> > > > Luciano Resende
> > > > http://twitter.com/lresende1975
> > > > http://lresende.blogspot.com/
> > >
> >
> >
> >
> > --
> > *Madison J. Myers*
> > *UC Berkeley, Master of Information & Data Science '17*
> >
> > *King's College London, MA Political Science '14*
> > *New York University, BA Political Science '12*
> >
> >    -
> >       LinkedIn <http://linkedin.com/in/madisonjmyers>
> >
>
>
>
> --
> Luciano Resende
> http://twitter.com/lresende1975
> http://lresende.blogspot.com/
>



-- 
*Madison J. Myers*
*UC Berkeley, Master of Information & Data Science '17*

*King's College London, MA Political Science '14*
*New York University, BA Political Science '12*

   -
      LinkedIn <http://linkedin.com/in/madisonjmyers>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message