FR 27.3. -- Short Course Machine Learning -- NOT canceled!

[Madlene Nussbaum]

Dear all

Everything is canceled nowadays. This is not the case for the scheduled machine learning
course!

It takes just a different format.

Please join in as you like!
On the Doodle list were 20 people, so I might not be all alone in front of a screen.

Happy home office + take care
Madlene



======= Program -- Distance Learning =======

**Short Course: Mastering Machine Learning for Prediction**

When: FR 27.3.2020, chat from 9-17, video calls 13:00 and 16:00
Online, over MS Teams, see below [1]


Language video calls: English if requested, else German.
Language videos and materials: English.


Expected previous knowledge:
Basic know-how of simple and multiple linear regression
(ordinary least squares-method).


Teaching Materials:
https://github.com/mnocci/2019_OpenGeoHub_machine-learning-madlene

Get everything: Green Button: Clone or Download > Download ZIP

Software requirements:
Microsoft Teams, R, e.g. RStudio, some R packages, see below [2]


------------------------------

Schedule
------------------------------

Morning
========

** Repeat linear regression if needed (self-study):

e.g. catch up by watching these very simple explanatory videos (1 hour on speed 1.25):
https://www.youtube.com/watch?v=Qa2APhWjQPc
https://www.youtube.com/watch?v=iAgYLRy7e20
https://www.youtube.com/watch?v=dQNpSa-bq4M


** Lecture Part I (self-study, 1:15 h):

Introduction and Overview of Machine Learning Methods for Prediction,
Overfitting. Methods in the Context of Supervised Learning: Lasso, Support Vector
Machines, Random Forest, Boosting, Model averaging.

https://www.youtube.com/watch?v=2pdRk4cj1P0


** Exercises on Part I (self-study, 1-2 h)

R-Code: exercises/OpenGeoHub-machine-learning-training-1.R
PDF:    exercises/OpenGeoHub-machine-learning-training-1.pdf


Afternoon
===========

** Video call session -- 13:00 --

1) Questions on linear regression
2) Questions on lecture I
3) Questions on exercises I
4) Brief discussion of exercises I



** Lecture Part II (self-study):

Model Selection and Model Building, Model Interpretation, Prediction
Intervals, Uncertainty, parametric and non-parametric Bootstrap

https://www.youtube.com/watch?v=Xq8b7hJa7GU (up to min 55:00 h)


** Exercises on Part II (self-study 1 h)

R-Code: exercises/OpenGeoHub-machine-learning-training-2.R
PDF:    exercises/OpenGeoHub-machine-learning-training-2.pdf

Brief discussion of exercises: rest of video part I above.


** Video call session -- 16:00 --

1) Questions on lecture II
2) Questions on exercises II
3) General questions on your data and your problems


** Lecture Part III: Model Evaluation, Cross-Validation, Validation Statistics, Good
Practice and Interpretation, R2 vs. r2, Scores for Classification.

not yet ready, sorry - will be delivered later.


-----



[1] Video Call and Chat

Please link into this Teams session (you click the link already now and hang up directly).
I will be on video call the given hours, on chat during whole Friday and respond on the
chat conversation sporadically before Friday (for e.g. installation problems).

MSTeams-Knigge:
* Please don't be shy to use the chat during the whole day!
* Please indicate in the chat that you have a question during the video conference, then I
can easily moderate.

Go to a web browser and paste the link:

https://teams.microsoft.com/l/meetup-join/19%3ameeting_ODA2Yjg2ZjktNjk0YS00YjU5LWIzMzAtODdlMTU2MzljNmM1%40thread.v2/0?context=%7b%22Tid%22%3a%22d6a1cf8c-768e-4187-a738-b6e50c4deb4a%22%2c%22Oid%22%3a%220233fdc4-2700-44e2-9a54-6a7562c45893%22%7d



[2] Installation:

You need R and an Editor, e.g. RStudio

R e.g. for Windows (the version of the software center is very old):
https://cran.r-project.org/ > Download R for Windows > base > Download R 3.6.3 for Windows

Rstudio:
https://rstudio.com/products/rstudio/download/#download

Following packages, run R code:
install.packages(c("grpgreg", "glmnet", "kernlab", "caret", "randomForest", "mboost",
                   "gbm", "geoGAM", "raster", "quantregForest"))


-- 

Madlene Nussbaum
___________________________________________________________

Telefon mobil    +41 79 761 34 66
madlene.nussbaum at bfh.ch