- Averaged Perceptron (bolt.trainer.avgperceptron.AveragedPerceptron)
Multi-class classification using the averaged perceptron by Freund and Schapire.
- Maximum Entropy (bolt.trainer.maxent.MaxentSGD)
Multinomial logistic regression classifier using SGD. Supports probabilistic output.
- Windows compatibility
Tested on Windows 7 using Enthought Python Distribution.
- Elastic-net penalty
Sparse regularization of correlated predictors.
- PEGASOS (bolt.trainer.sgd.PEGASOS)
New SVM solver based on projected (sub-)gradients.
- Generalized Linear Models (bolt.model.GeneralizedLinearModel)
Generalization of linear models to multi-class classification.
- OVA (bolt.trainer.OVA)
A one-versus-all approach to multi-class classification.
- Profiling of Cython code
Substantial performance improvement through profiling of Cython code.
- Major refactoring
changed the structure of the sub-modules (i.e., moved the SGD code to the bolt.trainer module) and introduced the bolt.io.Dataset interface in bolt.io.
- Cross-validation script added.
- Bugfix release
- Initial Release