An introduction and hands-on demo
||Presentation (Powerpoint & PDF)
||Demo - main script (bash)
|Random data-set generator (perl)
|Chart density of a single data column (R+ggplot2)
|Chart one numeric column against another (R+ggplot2)
This is an introduction to supervised machine learning, through
a presentation and hands-on demo
of one of the best open-source machine learning tools available today:
In this talk we'll learn:
- What exactly is machine learning?
- Can we make computers learn from data?
- What's the link between machine learning and optimization?
- When we should use machine learning instead of writing code?
- What makes vowpal wabbit such a great tool for machine learning?
- How can we automatically separate the signal from the noise?
- Will Elmer Fudd befuddle Bugs, or will the damn wabbit win again?
To run the demo, make sure you have the following prerequisites on your machine:
- vowpal-wabbit (vw)
- R + ggplot2
Download the 4 pieces of code from the table above, make sure
they are all in the current directory & executable:
chmod +x vw-demo random-poly distrib.r x-vs-y.r
The whole demo is scripted, so you just need to read the next
command step explanation and hit [enter] repeatedly to run the
Developed and tested on Ubuntu Linux.
Should work whereever
bash + R + vw are available.
You may need to change
'gwenview' on non Linux envs to any utility that can
Aug 26, 2013 update:
The "tell signal from noise" demo is now an integral
part of John Langford's
vowpal wabbit master source-tree
2015: statically linked precompiled vw binaries for Linux IA64