RAMP challenges begin with an interesting supervised prediction
problem for which there is data. Domain experts teaching
clean and curate the data and formalize the problem, working
with experienced data scientists if required. This process can
take two weeks to six months, and results in a 'starting kit',
which typically consists of:
-
A jupyter notebook that introduces the predictive
problem, provides some background information, exploratory
data analysis and data visualisation, explains the workflow
and provides a simple example solution.
-
A 'problem.py' file that parametrizes the setup and uses
building blocks from
RAMP workflow.
-
The data, split into training and test subsets.
-
A requirements file detailing all the required software
packages.
A challenge event can now be organised. These events may
be part of a post-graduate level programme or a 'hackathon'.
If the data science problem requires the mastering
of a specific tool, the RAMP event can be preceded by a
Training Sprint for explaining specific tools to the participants.
RAMP data challenges has been used for teaching in many courses
at various levels. Students usually love the platform for:
- its competitive gaming aspect,
- being able to learn from each other in the open phase, and
- to compete as a group with other student groups on the same problem.
Using RAMP for teaching
If you are interested in using the platform for teaching,
we can open a new event for any
RAMP challenge. You will have full control over the timeline
of the event (start, moving from closed to open phase and end).
At the end of your challenge we will give you a csv file with all
the performance metrics of each student.
Creating a new RAMP challenge
If you have a predictive problem you may be interested in
developing a new RAMP challenge to incite data scientists to
collaboratively prototype new models. All RAMP packages are
open source and documented, allowing you to build your challenge using
ramp-workflow and run an event by setting up your own RAMP
server using
ramp-board.
We may be able to work with you in developing your RAMP
challenge or helping to organise events such as hackathons
or use the problem in a classroom setting. We can also
automatically benchmark the thousands of models that are already
in the platform.
Contact us
To use RAMP for teaching or to work with us to develop a new
RAMP challenge or organise a new event, contact us at
admin@ramp.studio.