Models for predicting similarity between exemplars

摘要

An exemplar dictionary is built from exemplars of digital content for determining predictor blocks for encoding and decoding digital content. The exemplar dictionary organizes the exemplars as clusters of similar exemplars. Each cluster is mapped to a label. Machine learning techniques are used to generate a prediction model for predicting a label for an exemplar. The prediction model can be a hashing function that generates a hash key corresponding to the label for an exemplar. The prediction model learns from a training set based on the mapping from clusters to labels. A new mapping is obtained that improves a measure of association between clusters and labels. The new mapping is used to generate a new prediction model. This process is repeated in order to iteratively refine the machine learning modes generated.

主权项

1. A computer-implemented method for generating a prediction model for predicting clusters of similar exemplars for a target block of digital content, the method comprising:
receiving a set of labels and a set of clusters, each cluster comprising similar exemplars of digital content; generating an initial mapping from the set of clusters to the set of labels, the mapping assigning a label to each cluster; and generating the prediction model for predicting clusters of similar exemplars for a target block of digital content, the generating comprising iteratively:
training the prediction model using the labels assigned to the clusters of exemplars;predicting labels for exemplars of the clusters using the trained prediction model;determining strengths of association between the labels assigned to the clusters and labels predicted using the trained prediction model; andgenerating a new mapping from the same set of clusters to the same set of labels by reassigning labels to clusters based on the strengths of association between the assigned labels and the predicted labels.