公开(公告)号：US20200175379A1

公开(公告)日：2020-06-04

申请号：US16699727

申请日：2019-12-01

Applicant: Tsinghua University

Inventor： Huaqiang Wu ,  Bin Gao ,  Yudeng Lin ,  He Qian

IPC: G06N3/08 ,  G06N3/06 ,  G06N3/04

Abstract: A generative adversarial network device and a training method thereof. The generative adversarial network device includes a generator and a discriminator. The generator is configured to generate a first sample according to an input data; the discriminator is coupled to the generator, and is configured to receive the first sample and be trained based on the first sample; the generator includes a first memristor array serving as a first weight array. The generative adversarial network device can omit a process of adding noise to fake samples generated by the generator, thereby saving training time, reducing resource consumption and improving training speed of the generative adversarial network.

公开(公告)号：US11574199B2

公开(公告)日：2023-02-07

申请号：US16699727

申请日：2019-12-01

Applicant: Tsinghua University

Inventor： Huaqiang Wu ,  Bin Gao ,  Yudeng Lin ,  He Qian

IPC: G06N3/08 ,  G06N3/04 ,  G06N3/06

Abstract: A generative adversarial network device and a training method thereof. The generative adversarial network device includes a generator and a discriminator. The generator is configured to generate a first sample according to an input data; the discriminator is coupled to the generator, and is configured to receive the first sample and be trained based on the first sample; the generator includes a first memristor array serving as a first weight array. The generative adversarial network device can omit a process of adding noise to fake samples generated by the generator, thereby saving training time, reducing resource consumption and improving training speed of the generative adversarial network.