Propagation of distributions by Monte-Carlo sampling: Real number types with uncertainty represented by samples.

Code and website for DAL (Discriminative Active Learning) - a new active learning algorithm for neural networks in the batch setting. For the blog:

SAND: Semi-Supervised Adaptive Novel Class Detection and Classification over Data Stream

ECHO is a semi-supervised framework for classifying evolving data streams based on our previous approach SAND. The most expensive module of SAND is the change detection module, which has cubic time complexity. ECHO uses dynamic programming to reduce the time complexity. Moreover, ECHO has a maximum allowable sliding window size. If there is no concept drift detected within this limit, ECHO updates the classifiers and resets the sliding window. Experiment results show that ECHO achieves significant speed up over SAND while maintaining similar accuracy. Please refer to the paper (mentioned in the reference section) for further details.

PHD Thesis at CERFACS: Uncertainty Quantification for High Dimensional Problems

Open Active Learning for Deep Models