Glow:具有可逆1x1卷积的生成流
Glow: Generative Flow with Invertible 1x1 Convolutions
摘要:
基于流的生成模型(Dinh等人,2014)由于精确对数似然的易处理性、精确潜变量推断的易处理性以及训练和合成两者的并行性,在概念上是有吸引力的。在本文中,我们提出了Glow,这是一种使用可逆1x1卷积的简单生成流。使用我们的方法,我们证明了一个显着的改善,在标准基准的对数似然。也许最引人注目的是,我们证明了一个生成模型优化的平原对数似然目标是能够有效的逼真的合成和操作的大图像。我们模型的代码可在https://github.com/openai/glow上获得
英文摘要:
Flow-based generative models (Dinh et al., 2014) are conceptually attractive due to tractability of the exact log-likelihood, tractability of exact latent-variable inference, and parallelizability of both training and synthesis. In this paper we propose Glow, a simple type of generative flow using an invertible 1x1 convolution. Using our method we demonstrate a significant improvement in log-likelihood on standard benchmarks. Perhaps most strikingly, we demonstrate that a generative model optimized towards the plain log-likelihood objective is capable of efficient realistic-looking synthesis and manipulation of large images. The code for our model is available at https://github.com/openai/glow
机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。