Disclosed is a method for training a class conditional generative adversarial sequence network. A computer-implemented method and system for training a class conditional generative adversarial network(GAN) is described. The discriminator is trained using a classification loss function, omitting the use of an adversarial loss function. Alternatively, if the training data has C classes, the classification loss function may be expressed as a 2C class classification problem by which the discriminator is trained to distinguish the C classes twice. It is shown that such a trained discriminator provides an informative training signal to a generator to learn class condition data synthesis by the generator. Further, a data synthesis system and a computer-implemented method for synthesizing data using a generated portion of a trained generative adversarial network are described.
训练类条件生成对抗序列网络。描述了一种用于训练类条件生成对抗网络(GAN)的计算机实现的方法和系统。使用分类损失函数训练鉴别器,而省略使用对抗损失函数。取而代之地,如果训练数据具有C个类,则分类损失函数可以被表述为2C类分类问题,通过所述2C类分类问题,鉴别器被训练为对C个类区分2次。示出了,这样的经训练的鉴别器为生成器提供信息性训练信号,以通过生成器学习类条件数据合成。此外,描述了一种用于使用经训练的生成对抗网络的生成部分来合成数据的数据合成系统和计算机实现的方法。
TRAINING A CLASS-CONDITIONAL GENERATIVE ADVERSERIAL NETWORK
训练类条件生成对抗序列网络
2021-03-26
Patent
Electronic Resource
Chinese
IPC: | G06K Erkennen von Daten , RECOGNITION OF DATA / B60W CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION , Gemeinsame Steuerung oder Regelung von Fahrzeug-Unteraggregaten verschiedenen Typs oder verschiedener Funktion / G06N COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS , Rechnersysteme, basierend auf spezifischen Rechenmodellen |