Skip to content
/ gmdl Public

Classifier based on the Minimum Description Length Principle

License

Notifications You must be signed in to change notification settings

brenolf/gmdl

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

84 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Gaussian Mixture Descriptors Learner (GMDL)

Install

Dependencies

Run sh ./install.sh. On Ubuntu-based systems run this command as sudo.

App

Run make and ./gmdl.app to run.

Usage

usage: GMDL [options] ... 
options:
  -i, --inline               pass configurations parameters inline instead of reading json file
      --stdin                reads from stdin (prioritized)
  -q, --quiet                omits logging when classification fails
      --online               makes algorithm online
      --confusion-matrix     prints confusion matrix
      --label                the index of the column that should be considered the label (int [=0])
      --labels               the labels of the dataset comma-separated (string [=])
      --set                  the key to the set in the config file that should be read (string [=])
  -p, --path                 the path in which to look for datasets (string [=])
      --training             the name of the traning set (string [=])
      --testing              the name of the testing set (string [=])
      --config               the path to the config json file (string [=./config.json])
      --learning_rate        the learning rate used to the SGD on the theta exponents (double [=0.01])
      --momentum             the momentum used to the SGD on the theta exponents (double [=0.9])
      --tau                  the exponent of the prototype distance used to separate classes (double [=1])
      --omega                the default description length assumed when there is no clue about the data being assessed (double [=-32])
      --forgetting_factor    the factor by which the samples in the mixture are considered outdated (1 = no forgetting) (double [=1])
      --sigma                the standard deviation to the noise applied when the covariance matrix is getting close to be singular (double [=1])
      --dimension            the number of attributes in the dataset (mandatory for online learning) (int [=0])
  -?, --help                 print this message

You can also use a config.json with the same keys as in config.example.json. Note that you must use this json in order to provide the classes being examined.

About

Classifier based on the Minimum Description Length Principle

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages