Towards Understanding the Invertibility of Convolutional Neural Networks
Abstract
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.
Cite
Text
Gilbert et al. "Towards Understanding the Invertibility of Convolutional Neural Networks." International Joint Conference on Artificial Intelligence, 2017. doi:10.24963/IJCAI.2017/236Markdown
[Gilbert et al. "Towards Understanding the Invertibility of Convolutional Neural Networks." International Joint Conference on Artificial Intelligence, 2017.](https://mlanthology.org/ijcai/2017/gilbert2017ijcai-understanding/) doi:10.24963/IJCAI.2017/236BibTeX
@inproceedings{gilbert2017ijcai-understanding,
title = {{Towards Understanding the Invertibility of Convolutional Neural Networks}},
author = {Gilbert, Anna C. and Zhang, Yi and Lee, Kibok and Zhang, Yuting and Lee, Honglak},
booktitle = {International Joint Conference on Artificial Intelligence},
year = {2017},
pages = {1703-1710},
doi = {10.24963/IJCAI.2017/236},
url = {https://mlanthology.org/ijcai/2017/gilbert2017ijcai-understanding/}
}