上周一開始閱讀了香港中文大學(xué)Bo Dai等人在ICCV 2017的文章想罕,Towards Diverse and NaturalImage Des...
這個(gè)工作著眼于Visual Question Generation (VQG)這個(gè)多模態(tài)問題,同樣使用當(dāng)下熱門的GAN來(lái)處理霉涨,它最大的創(chuàng)新之處在...
漸漸體會(huì)到GAN訓(xùn)練難確實(shí)是一個(gè)讓人頭疼的問題按价,一個(gè)多月前我曾粗略地了解了一下WGAN,知道這是一個(gè)著眼于提高GAN訓(xùn)練穩(wěn)定性的成果笙瑟,但后來(lái)發(fā)現(xiàn)...
上周讀了CVPR 2017一篇名為DeLiGAN : GenerativeAdversarial Networks for Diverse an...
CAAE即Conditional Adversarial Autoencoder楼镐,這個(gè)概念最早來(lái)自論文Age Progression/Regre...
DP-GAN指的是Diversity-Promoting GAN(而不是另外一篇論文里提出的DifferentiallyPrivateGAN )...
雖然前段時(shí)間稍微了解過Policy Gradient,但后來(lái)發(fā)現(xiàn)自己對(duì)其原理的理解還有諸多模糊之處往枷,于是希望重新梳理一番框产。Policy Grad...
SeqGAN的概念來(lái)自AAAI 2017的SeqGAN: Sequence Generative Adversarial Nets with P...
InfoGAN的概念最早來(lái)自論文InfoGAN: Interpretable Representation Learning by Inform...