Towards Understanding the Invertibility of Convolutional Neural Networks
Several recent works have empirically observed that Convolutional Neural Nets (CNNs) are (approximately) invertible. To understand this approximate invertibility phenomenon and how to leverage it more effectively, we focus on a theoretical explanation and develop a mathematical model of sparse signal recovery that is consistent with CNNs with random weights. We give an exact connection to a particular model of model-based compressive sensing (and its recovery algorithms) and random-weight CNNs. We show empirically that several learned networks are consistent with our mathematical analysis and then demonstrate that with such a simple theoretical framework, we can obtain reasonable re- construction results on real images. We also discuss gaps between our model assumptions and the CNN trained for classification in practical scenarios.
Code (0)
등록된 구현이 없습니다.
Tasks
Compressive SensingGeneral ClassificationSimilar Papers 제목 키워드 기반
Neural Networks Preserve Invertibility Across Iterations: A Possible Source of Implicit Data Augmentation
Determining what kind of representations neural networks learn, and how this may relate to generalization, remains a challenging problem. Previous work has utilized a rich set of methods to invert layer representations o…
Data AugmentationInvertibility of Convolutional Generative Networks from Partial Measurements
In this work, we present new theoretical results on convolutional generative neural networks, in particular their invertibility (i.e., the recovery of input latent code given the network output). The study of network inv…
Image InpaintingProduction Function Estimation without Invertibility: Imperfectly Competitive Environments and Demand Shocks
We advance the proxy variable approach to production function estimation. We show that the invertibility assumption at its heart is testable. We characterize what goes wrong if invertibility fails and what can still be d…
Analysis of Invariance and Robustness via Invertibility of ReLU-Networks
Studying the invertibility of deep neural networks (DNNs) provides a principled approach to better understand the behavior of these powerful models. Despite being a promising diagnostic tool, a consistent theory on their…
DiagnosticUnderstanding and Mitigating Exploding Inverses in Invertible Neural Networks
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 fr…