DeepAI AI Chat
Log In Sign Up

Understanding and mitigating exploding inverses in invertible neural networks

06/16/2020
by   Jens Behrmann, et al.
14

Invertible neural networks (INNs) have been used to design generative models, implement memory-saving gradient computation, and solve inverse problems. In this work, we show that commonly-used INN architectures suffer from exploding inverses and are thus prone to becoming numerically non-invertible. Across a wide range of INN use-cases, we reveal failures including the non-applicability of the change-of-variables formula on in- and out-of-distribution (OOD) data, incorrect gradients for memory-saving backprop, and the inability to sample from normalizing flow models. We further derive bi-Lipschitz properties of atomic building blocks of common architectures. These insights into the stability of INNs then provide ways forward to remedy these failures. For tasks where local invertibility is sufficient, like memory-saving backprop, we propose a flexible and efficient regularizer. For problems where global invertibility is necessary, such as applying normalizing flows on OOD data, we show the importance of designing stable INN building blocks.

READ FULL TEXT

page 1

page 4

page 5

page 19

page 20

page 21

page 23

page 29

10/08/2018

Algorithmic Aspects of Inverse Problems Using Generative Models

The traditional approach of hand-crafting priors (such as sparsity) for ...
05/13/2021

Provably Convergent Algorithms for Solving Inverse Problems Using Generative Models

The traditional approach of hand-crafting priors (such as sparsity) for ...
07/18/2019

MintNet: Building Invertible Neural Networks with Masked Convolutions

We propose a new way of constructing invertible neural networks by combi...
02/21/2022

On the Suitability of Neural Networks as Building Blocks for The Design of Efficient Learned Indexes

With the aim of obtaining time/space improvements in classic Data Struct...
01/18/2023

Discrete Latent Structure in Neural Networks

Many types of data from fields including natural language processing, co...

Code Repositories

INN-exploding-inverses

Code for Understanding and Mitigating Exploding Inverses in Invertible Neural Networks http://arxiv.org/abs/2006.09347


view repo