Tutorials

Geocomputing with R

https://github.com/jannes-m/erum18_geocompr

https://geocompr.robinlovelace.net/

Efficient R programming

https://github.com/csgillespie/efficient

https://csgillespie.github.io/efficientR/

Reproducible data screening and quality control

http://www.biostatistics.dk/eRum2018/

https://ekstroem.github.io/dataMaid/

Building a package that lasts

Sources

Tuesday

Machine learning

Source: Twitter

GBM better than deep lerning

Benchmarks

GBM benchmarks

RStudio:

• Practitioners should use Keras library

• Use R package tfruns for logging model performance progress

• If training takes longer than > 5min, cloudml could be an option https://tensorflow.rstudio.com/tools/cloudml/articles/getting_started.html

https://github.com/andrie/erum-tensorflow-tools

Slides of other talks:

• Show me your model 2.0, R Package DALEX; alternative R package lime

• Modelling Item Worth Based on Rankings

• Faster package installation

• R/shiny interface for interactive visualization of data in objects derived from the SummarizedExperiment class

• Errors & conditions

• First steps with Stan

Wednesday

• Keynote: History of R

• Data science: stories instead of code, thinking of the long-term horizon (value of data!)

• Building a data science team around R

  • Infrastructure questions (How to access R / get packages)

• Drilldown data discovery with shiny