PLEASD stands for Prediction and LEArning for Structured Data. It is a Matlab toolbox of algorithmic frameworks for training structured prediction models. We provide this toolbox to ease the process of applying structured learning to new problems. We attempt to minimize users’ involvement in coding the structured learning framework such that they can focus on issues related to their specific problems. Currently, PLEASD has included the following structured learning frameworks:
1. Bundle method for risk minimization;
2. Structured perceptron learning;
3. Structured learning from partial annotations;
4. Structured perceptron learning from partial annotations.
Please forward any suggestions, bug reports, questions to firstname.lastname@example.org. Your feedback is highly appreciated.
(c) MIT License for worry-free use and distribution.
09/11/2012: fixed some errors in the User Guide -> search_space is determined by d.y_dot (the annotation), not d.y_hat (the prediction).