markov thebeast is a Markov Logic interpreter. We also see it as structured prediction framework in which the user can define a loglinear distribution over a complex output space.
thebeast provides MAP inference using a Cutting Plane based technique that was shown to be very efficient (see paper). Parameter Estimation is performed via Online Learning and 3 different update rules: MIRA, Perceptron and Passive-Aggressive-Learning.
thebeast comes with an interactive shell that allows the user to inspect the trained model in various ways.
Version 0.0.2 comes with an improved manual, an additional example application and some small bugfixes.
- Changes to previous version:
Initial Announcement on mloss.org.
No one has posted any comments yet. Perhaps you'd like to be the first?
Leave a comment
You must be logged in to post comments.