Analyzing Inverse Problems with Invertible Neural Networks
By Lynton Ardizzone, Jakob Kruse, Sebastian J. Wirkert, D. Rahner, E. Pellegrini, Ralf Klessen, Lena Maier-Hein, Carsten Rother, Ullrich Köthe. 2018
598 citations recorded by the source corpus.
In many tasks, in particular in natural science, the goal is to determine hidden system parameters from a set of measurements. Often, the forward process from parameter- to measurement-space is a well-defined function, whereas the inverse problem is ambiguous: one measurement may map to multiple different sets of parameters. In this setting, the posterior parameter distribution, conditioned on an input measurement, has to be determined. We argue that a particular class of neural networks is well suited for this task -- so-called Invertible Neural Networks (INNs). Although INNs are not new, they have, so far, received little attention in literature. While classical neural networks attempt to solve the ambiguous inverse problem directly, INNs are able to learn it jointly with the well-defined forward process, using additional latent output variables to capture the information otherwise lost. Given a specific measurement and sampled latent variables, the inverse pass of the INN provides a full distribution over parameter space. We verify experimentally, on artificial data and real-world problems from astrophysics and medicine, that INNs are a powerful analysis tool to find multi-modalities in parameter space, to uncover parameter correlations, and to identify unrecoverable parameters.
Subfields: Artificial Intelligence, Statistical and Nonlinear Physics, Neural Architecture Search & AutoML, AI for Physical Sciences
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
- Sergey Levine
- Yoshua Bengio
- Ruslan Salakhutdinov
- Mohit Bansal
- Quoc V. Le
- Pieter Abbeel
- Percy Liang
- Trevor Darrell
In Statistical and Nonlinear Physics
- David Kristjanson Duvenaud
- Jörn-Henrik Jacobsen
- Jens Behrmann
- Lars Ruthotto
- Eldad Haber
- Arnaud Doucet
- Ullrich Köthe
- Cho-Jui Hsieh
In Neural Architecture Search & AutoML
In AI for Physical Sciences
Citations in the corpus
- Flexibly Fair Representation Learning by Disentanglement (2019, background): “Designing invariant representations with non-variational objectives has also been explored, including in reversible models (Ardizzone et al., 2019; Jacobsen et al., 2019).”
- Excessive Invariance Causes Adversarial Vulnerability (2018, methodology): “We also leverage recent advances in reversible (Gomez et al., 2017) and bijective networks (Jacobsen et al., 2018; Ardizzone et al., 2018; Kingma & Dhariwal, 2018) for our analysis.”
- Invertible Residual Networks (2018, background): “The ability to invert deep networks alleviates the need to store intermediate activations during backpropagation [11, 6] and offers a promising approach in Inverse Problems [3].”
- Augmented Neural ODEs (2019, background): “(2018) also augment the input with zeros to build invertible ResNets and transformations.”
- Parrot: Data-Driven Behavioral Priors for Reinforcement Learning (2020, background): “(2017) allows using deep neural networks to parameterize this mapping (making it expressive) While the original real NVP work modelled unconditional distributions, follow-up work has found that it can be easily extended to incorporate conditioning information (Ardizzone et al., 2019).”
- Surgical data science – from concepts toward clinical translation (2020, background): “Even if a sample is in the support of the training distribution, a problem may not be uniquely solvable (Ardizzone et al., 2018) or the solution may be associated with high uncertainty.”
- CFLOW-AD: Real-Time Unsupervised Anomaly Detection with Localization via Conditional Normalizing Flows (2021, methodology): “The code is in PyTorch [28] with the FrEIA library [1] used for generative normalizing flow modeling.”
- Stochastic Image-to-Video Synthesis using cINNs (2021, background): “[1] Lynton Ardizzone, Jakob Kruse, Carsten Rother, and Ullrich Köthe.”
- BayesFlow: Learning Complex Stochastic Models With Invertible Neural Networks (2020, methodology): “Following [2], we construct a scenario in which the observed data x is a one-hot encoded vector representing one of the labels red, green, blue, or yellow (hard assignment of labels).”
- Guided Image Generation with Conditional Invertible Neural Networks (2019, methodology): “We propose a third approach, by extending Invertible Neural Networks (INNs, [8, 22, 1]) for the task of conditional image generation, by adding conditioning inputs to their core building blocks.”
- Making Sense of CNNs: Interpreting Deep Representations & Their Invariances with INNs (2020, methodology): “Previous works have successfully applied INNs for density estimation [12], inverse problems [2], and on top of autoencoder representations [15,61].”
- An introduction to deep generative modeling (2021, background): “Both of these approaches belong to the more general class of invertible neural networks; see, for example, the excellent literature review and application to inverse problems in Reference [1].”
- Neural SDE: Stabilizing Neural ODE Networks with Stochastic Noise (2019, methodology): “A memory efficient approach to compute the gradients by adjoint methods was developed, making it possible to train large, multi-scale generative networks [10, 11].”
- iPOKE: Poking a Still Image for Controlled Stochastic Video Synthesis (2021, background): “Invertible neural networks (INNs) are learnable bijective functions often used to transform between two probability distributions, thus being a natural choice for addressing inverse problems [3], introspecting and explaining neural network representations [22, 32] and domain transfer [62, 63].”
- CDDFuse: Correlation-Driven Dual-Branch Feature Decomposition for Multi-Modality Image Fusion (2022, background): “The invertible neural network is an important module of the Normalized Flow model, a popular kind of generative model [2].”
- Hierarchical Conditional Flow: A Unified Framework for Image Super-Resolution and Image Rescaling (2021, background): “Flow-based models [7, 8, 23, 37, 13, 22, 2, 16, 12, 36, 28] aim to learn a bijective mapping between the target space and the latent space.”