build your own data challenge, or just organize team work
TRANSCRIPT
RAMP-WORKFLOW & RAMP-KITS
1
Université Paris-Saclay / CNRSBALÁZS KÉGL
Center for Data ScienceParis-Saclay
2
Most classical data challenges are HR and publicity events
3
We decided to turn them into a tool for
1. Collaborative prototyping 2. Teaching aid 3. Data science team management
4
We are open sourcing it
toolkit: https://github.com/paris-saclay-cds/ramp-workflow
examples: https://github.com/ramp-kits
5
Funded by Université Paris-Saclay
6
RAMP.STUDIO DATA CHALLENGE WITH CODE SUBMISSION
Center for Data ScienceParis-SaclayB. Kégl (CNRS) 7
what you achieved with a well tuned deep net
the diversity gap
the human blender gap
competitive phase
collaborative phase
THE POWER OF THE (COLLABORATING) CROWD
Center for Data ScienceParis-SaclayB. Kégl (CNRS)
OPEN PHASE LETS PARTICIPANTS CATCH UP THE GOAL OF TEACHING
8
Center for Data ScienceParis-SaclayB. Kégl (CNRS) 9
COMMUNICATION AND REUSE
10
You can
1. Participate in upcoming RAMPs 2. Use RAMP in teaching or training
11
Setting up the RAMP is was
long and hard.
12
Separate workflow building and workflow optimization
13
Before solving the problem, set it up
(even put it into production)
• toolkit: https://github.com/paris-saclay-cds/ramp-workflow
• for designing workflows
• set of ready-made metrics, workflows, CV schemes, data readers
• unique command-line test script
• examples: https://github.com/ramp-kits
• a zoo of problems, experiments, workflows
• (at least) one initial solution
14
RAMP-WORKFLOW & RAMP-KITS
Center for Data ScienceParis-SaclayB. Kégl (CNRS)
CLASSIFYING AND REGRESSING ON MOLECULAR SPECTRA
15
chemotherapy drug in elastic pocket
laser spectrometer
molecular spectra
feature extractor 1
feature extractor 2
regressor
concentration
classifier
drug type
Center for Data ScienceParis-SaclayB. Kégl (CNRS)
FORECASTING EL NINO SIX MONTHS AHEAD
16
… 300.14 299.83 298.76 299.87 299.82 300.15 300.10 299.50… …
time series feature extractor
x (a fixed length feature vector)regressor
17
A SINGLE SCRIPT TO DEFINE THE BUNDLE
X ypred score type
score
cross-validation scheme
data
con
nect
ors
FE CLF
workflow
18
A SINGLE EXECUTABLE TO TEST THE SUBMISSIONS
• Keep your different submissions in a simple file structure
• Communicate them on git
• Execute them also from the notebook
19
You can
1. Use rampwf for your own workflows 2. Use rampwf to organize workflow
building and optimization in an internal data science team
3. Submit it to us if you want to run a data challenge
20
toolkit: github.com/paris-saclay-cds/ramp-workflow
examples: github.com/ramp-kits
blogs: medium.com/@balazskegl
slack: ramp-studio.slack.com
frontend: www.ramp.studio
mail: [email protected]