The ability of Variational Autoencoders (VAEs) to learn disentangled representations has made them popular for practical applications. However, their behaviour is not yet fully understood. For example, the questions of when they can provide disentangled representations, or suffer from posterior collapse are still areas of active research. Despite this, there are no layerwise comparisons of the representations learned by VAEs, which would further our understanding of these models. In this paper, we thus look into the internal behaviour of VAEs using representational similarity techniques. Specifically, using the CKA and Procrustes similarities, we found that the encoders' representations are learned long before the decoders', and this behaviour is independent of hyperparameters, learning objectives, and datasets. Moreover, the encoders' representations in all but the mean and variance layers are similar across hyperparameters and learning objectives.
翻译:变异自动编码器(VAEs)能够学习分解的表达方式,这使他们在实际应用中很受欢迎,但是,他们的行为还没有得到完全理解。例如,他们何时能够提供分解的表达方式,或何时会遭受后天崩溃的问题仍然是积极研究的领域。尽管如此,对于VAE所学的表达方式,并没有进行分层的比较,这将增进我们对这些模型的理解。在本文中,我们用代表性的相似性技术来审视VAEs的内部行为。具体地说,使用CKA和Procrustes相似性,我们发现编码器的表述方式是在解密器远前学习的,而这种行为与超参数、学习目标和数据集无关。此外,在超参数和学习目标方面,编码器的表达方式与平均值和差异层是相似的。