State-of-the-art for semi-supervised learning is slow to train and the performance is sensitive to the choices of the labeled data and hyper-parameter values . The FROST method trains up to an order of magnitude faster and is more robust than state-of theart methods . FROST’s capability to perform well when the composition of the unlabeled data is unknown; that is when theunlabeled data contain unequal numbers of each class and can contain out of-distribution examples that don’t belong to any of the training classes . High performance, speed of training, and insensitivity to hyperparameters makeFROST the most practical method for one-shot¬†training, make FROST the¬†most practical

Author(s) : Helena E. Liu, Leslie N. Smith

Links : PDF - Abstract

Code :

Keywords : frost - training - data - state - robust -

Leave a Reply

Your email address will not be published. Required fields are marked *