Controlling neural networks via energy dissipation

Moeller M, Moellenhoff T, Cremers D (2019)


Publication Type: Conference contribution

Publication year: 2019

Publisher: Institute of Electrical and Electronics Engineers Inc.

Book Volume: 2019-October

Pages Range: 3255-3264

Conference Proceedings Title: Proceedings of the IEEE International Conference on Computer Vision

Event location: Seoul, KOR

ISBN: 9781728148038

DOI: 10.1109/ICCV.2019.00335

Abstract

The last decade has shown a tremendous success in solving various computer vision problems with the help of deep learning techniques. Lately, many works have demonstrated that learning-based approaches with suitable network architectures even exhibit superior performance for the solution of (ill-posed) image reconstruction problems such as deblurring, super-resolution, or medical image reconstruction. The drawback of purely learning-based methods, however, is that they cannot provide provable guarantees for the trained network to follow a given data formation process during inference. In this work we propose energy dissipating networks that iteratively compute a descent direction with respect to a given cost function or energy at the currently estimated reconstruction. Therefore, an adaptive step size rule such as a line-search, along with a suitable number of iterations can guarantee the reconstruction to follow a given data formation model encoded in the energy to arbitrary precision, and hence control the model's behavior even during test time. We prove that under standard assumptions, descent using the direction predicted by the network converges (linearly) to the global minimum of the energy. We illustrate the effectiveness of the proposed approach in experiments on single image super resolution and computed tomography (CT) reconstruction, and further illustrate extensions to convex feasibility problems.

Involved external institutions

How to cite

APA:

Moeller, M., Moellenhoff, T., & Cremers, D. (2019). Controlling neural networks via energy dissipation. In Proceedings of the IEEE International Conference on Computer Vision (pp. 3255-3264). Seoul, KOR: Institute of Electrical and Electronics Engineers Inc..

MLA:

Moeller, Michael, Thomas Moellenhoff, and Daniel Cremers. "Controlling neural networks via energy dissipation." Proceedings of the 17th IEEE/CVF International Conference on Computer Vision, ICCV 2019, Seoul, KOR Institute of Electrical and Electronics Engineers Inc., 2019. 3255-3264.

BibTeX: Download