no code implementations • 20 Dec 2017 • Sebastian Sudholt, Gernot Fink
By taking a probabilistic perspective on training CNNs, we derive two different loss functions for binary and real-valued word string embeddings.
Attribute Segmentation +1