Volata
Volata is a generic machine learning system implemented in Java that provides ensemble methods like AdaBoost, Bagging, GloBoost and so on. Volata must be instanciated with generalization operators called lgg (least general generalization). Examples of such instanciations are given for classification tasks using attribute-value, sequences and trees representations of examples.
volata: VOtes of Least generAl generalizaTions in jAva.
Java archive available soon.
See some boosting simulations and animations produced using volata.
Description
- Main learners : DLG, GloBo, GloBoost, Bagging, AdaBoost.
- Positive Learners :
- on attribute-value: minimal rectangle
- on words: k-TSSI, ZR, REGEX
Volata applied to XML tasks
Document classification
- using attribute-value representation (GloBo+MinRect)
- using sequence representation (GloBo+TSSI)
Information extraction
- unary extraction from various organizations :
- n-ary extraction from various organizations :
Grammatical theories learned with volata
tic-tac-toe
- GloBo+TSSI (accuracy: 81.43 %)
- GloBo+ZR (accuracy: 95.84 %)
- GloBo+Fernau (accuracy: 99.77 %)
badges
- DLG+TSSI (accuracy: 73.61 %)
- AdaBoost+Fernau (accuracy: 100.00 %)
promoters
- DLG+ZR (accuracy: 50.00 %)
- GloBoost+Fernau (accuracy: 86.23 %)
splice-2classes
- AdaBoost+Fernau (accuracy: 94.91 %)
