hands-on training of data mining - european commission · introduction to data mining class baylor...
TRANSCRIPT
Hands-on Training of Data Mining Through Visual Programming and Interactive Data Analysis
Blaz Zupan University of Ljubljana, Slovenia
Introduction to Data Mining Class Baylor College of Medicine, Houston, September 2016
Introduction to Data Mining @ Mol Life of Stem Cells 2016 Ljubljana, September 2016
Data ScienceLinear algebra
Probability
Statistics
Mathematical optimization
Data representation
High performance computing
Scripting languages
Data visualization
Data Science for Data Owners
Intuition
No fuss with math or programming
Accessible & easy to use
Visualizations
Explorative interfaces
The Challenge
Consider data owners with no previous training or math, statistics or computer science.
Can we train them to reconnect with their data?
Can we do it in a single day?
Even with big data?
Interaction
Experimentation
On-The-Fly Data Creation
Do Some Crazy Stuff
ConclusionCan we train data owners to use data science?Yes.
Can we do it in two days?Most likely.
What is required?A good explorative analysis tool. And a good teacher.
Will data owners then become data scientists?No. Or, better, probably no.
Can this work on big data? Sure. No difference here.
How about deep learning and stuff? Embedding is low hanging fruit. Start here.
Thanks toGad Shaulsky Adam Kuspa Rafael Rosengarten
Funding ARRS
EU Comission NIH
Fulbright
Boro Nikić Marko Limbek
Janez Demšar Anže Starič
Aleš Erjavec Ajda Pretnar Tomaž Curk
Marko Toplak Tomaž Hočevar
Lan Žagar Vesna Tanko Jure Žbontar
Marinka Žitnik Martin Stražar Andrej Čopar