本文详细介绍了如何识别和诊断GAN（生成对抗网络）的失败模式，包括如何训练稳定的GAN，如何识别模式塌陷，以及如何判断收敛失败。模式塌陷表现为生成图像多样性低，而收敛失败则表现为判别器损失趋近于零。解决方案包括限制潜变量维度、检查数据量、调整模型容量和优化算法等。
GAN reading 1
How to Identify and Diagnose GAN Failure Modes
part 1: training a stable GAN
Training procedure:
- use the real image to train the discriminator
- generate the fake image and use the fake image to train the discriminator
- first, generate latent points (random vector/tensor)
- second, give latent points to the generator to generate the fake image
- third, use the fake image to train the discriminator
- generate latent points and labels
- use latent points and labels to train the GAN model and get the loss
- in the GAN model training process, discriminator training is disabled; so only the generat
最低0.47元/天 解锁文章
扫一扫
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
