Skip to yearly menu bar Skip to main content


Invertible Residual Networks

Jens Behrmann · Will Grathwohl · Ricky T. Q. Chen · David Duvenaud · Joern-Henrik Jacobsen

Pacific Ballroom #11

Keywords: [ Unsupervised Learning ] [ Supervised Learning ] [ Deep Generative Models ] [ Computer Vision ] [ Architectures ]


We show that standard ResNet architectures can be made invertible, allowing the same model to be used for classification, density estimation, and generation. Typically, enforcing invertibility requires partitioning dimensions or restricting network architectures. In contrast, our approach only requires adding a simple normalization step during training, already available in standard frameworks. Invertible ResNets define a generative model which can be trained by maximum likelihood on unlabeled data. To compute likelihoods, we introduce a tractable approximation to the Jacobian log-determinant of a residual block. Our empirical evaluation shows that invertible ResNets perform competitively with both state-of-the-art image classifiers and flow-based generative models, something that has not been previously achieved with a single architecture.

Live content is unavailable. Log in and register to view live content