News editors obtained the following quote from the background information suppli ed by the inventors:“The present invention relates generally to the fields of d eep learning models, efficient training of deeplearning models, and of classifi cation tasks performed by deep learning models.“To train a deep learning model to perform classification and then subsequently to attempt to add newclasses for the deep learning model can be problematic. Ad ding new classes may interfere with the abilityof the deep learning model to co rrectly classify the original classes. Moreover, adding new classes may beperfo rmed by accessing and supplementing the original training data but the original training data is notalways subsequently accessible. This inaccessibility may be due to data security and/or storage regulations.Moreover, repeating the traini ng with all of the original training data that is enhanced with new classes ist ime-consuming and resource-consuming.
NETL
NSTL