Gaussian Processes on Strings

shogun-toolbox / shogun

Shōgun

http://shogun-toolbox.org

BSD 3-Clause "New" or "Revised" License

3.03k stars 1.04k forks source link

Gaussian Processes on Strings #2072

Open karlnapf opened 10 years ago

karlnapf commented 10 years ago

A while ago, we aimed at making GPs in Shogun work with StringFeatures. I don't know the current state on this.

This task is to

make sure GPs can be used with String features
write a minimal example for python_modular
extend the existing GP notebook with a nice example, for example classifying amino-acid sequences with the Distance segments kernel in Shogun (ask me for details)

@votjakovr and me are able to answer questions here

pl8787 commented 10 years ago

@karlnapf What's the meaning of work with string features? Make the stirng feature as a binary vector? For example:

"a" -> [1,0,0]
"b" -> [0,1,0]
"c" -> [0,0,1]
"a" -> [1,0,0]

karlnapf commented 10 years ago

Have a look at the string features examples. The main point is that training data, rather than being a matrix of a certain type (with fixed dimensions), it rather is a collection of sequences of characters (might be any type), which might be of different length.