Skip to content

Releases: mlange-42/som

SOM v0.2.0

26 Jul 20:02
a35e7cf
Compare
Choose a tag to compare

SOM v0.2.0

Go implementation of Self-Organizing Maps (SOM) alias Kohonen maps.
Provides a command line tool and a library for training and visualizing SOMs.

Changelog

Features

  • Adds label propagation for semi-supervised learning (#43)
  • Adds switch --ignore to all prediction-related commands, for ignoring layers (#43)
  • Adds weight decay for better regularization during training (#44)
  • Can handle missing/no-data class labels (#45)
  • Adds CLI command fill for filling missing data (#49)
  • Adds CLI command predict for predicting complete layers (#50)
  • Adds codes plots showing a 2D array of small line, bar, pie, rose or image plot, one for each node (#52, #53, #53, #54, #55, #57, #61)
  • Adds CLI flag --sample to plot only a sub-sample of the input data (#56)
  • Draw optional class boundaries on all applicable plots (#60)
  • Adds command quality to calculate and print diverse SOM quality metrics (#62)
  • SOM parameters can be overwritten during training from CLI (#65)

Documentation

  • Examples are included in release downloads (#51)
  • Adds hand-written digits example (#55)
  • Adds world countries population pyramids example (#57)
  • Adds comprehensive help text for each CLI command (#58)
  • CLI help shows a command tree for commands with sub-commands (#58)

Bugfixes

  • Update BMU weights before the weights of neighboring nodes (#42)
  • Fix ViSOM weight update formula, with error in publication Yin 2002 (#65)

Other

  • All plots draw the legend on top of / after the actual plot (#48)
  • Improved CLI error messages, showing the affected command (#58)

SOM v0.1.0

21 Jul 21:24
bebb92d
Compare
Choose a tag to compare

SOM v0.1.0

Go implementation of Self-Organizing Maps (SOM) alias Kohonen maps.
Provides a command line tool and a library for training and visualizing SOMs.

Features

  • Multi-layered SOMs, alias XYF, alias super-SOMs.
  • Visualization Induced SOMs, alias ViSOMs.
  • Training from CSV files without any manual preprocessing.
  • Supports continuous and discrete data.
  • Fully customizable training and SOM parameters.
  • Visualization of SOMs through heatmaps and u-matrix plots, with data point labels.
  • Use as command line tool or as Go library.

Changelog

Initial release of SOM, the Self-organizing Maps library and CLI tool for Go.