Search Results for author:
Found 2 papers, 1 papers with code
United We Stand: Using Epoch-wise Agreement of Ensembles to Combat Overfit
Our method allows for the incorporation of useful knowledge obtained by the models during the overfitting phase without deterioration of the general performance, which is usually missed when early stopping is used.
The Dynamic of Consensus in Deep Networks and the Identification of Noisy Labels
This introduces a problem when training in the presence of noisy labels, as the noisy examples cannot be distinguished from clean examples by the end of training.
Cannot find the paper you are looking for? You can
Submit a new open access paper.
