Frac-Vector: Better Category Representation

For this paper, we proposed the fractional category representation vector (FV) based on calculus (FC), of which one-hot label is only special case when derivative order 0. FV can be considered as a distributional negative probability considered. FVs used either regularization method or distributed representation. They gain significantly in generalization classification models and representability generative adversarial networks with conditions (C-GANs). In image classification, linear combinations correspond to mixture images an independent variable loss function. Our experiments showed that also space sampling, fewer dimensions less computational overhead than normal distributions.