论文标题
邪教：基于典型的环境检测，持续无监督的学习
CULT: Continual Unsupervised Learning with Typicality-Based Environment Detection
论文作者
论文摘要
我们介绍了邪教（通过基于典型的环境检测的持续无监督的表示学习），这是一种通过各种自动编码器进行持续无监督学习的新算法。 Cult在VAE的潜在空间中使用一个简单的典型度量，以检测环境中的分布变化，该分布转移与生成性重播和一个辅助环境分类器一起使用，以限制在无限制的表示学习中的灾难性遗忘。在我们的实验中，邪教显着优于基线持续无监督的学习方法。本文的代码可以在此处找到：https：//github.com/oliveradk/cult
We introduce CULT (Continual Unsupervised Representation Learning with Typicality-Based Environment Detection), a new algorithm for continual unsupervised learning with variational auto-encoders. CULT uses a simple typicality metric in the latent space of a VAE to detect distributional shifts in the environment, which is used in conjunction with generative replay and an auxiliary environmental classifier to limit catastrophic forgetting in unsupervised representation learning. In our experiments, CULT significantly outperforms baseline continual unsupervised learning approaches. Code for this paper can be found here: https://github.com/oliveradk/cult
