Trainable Spectrally Initializable Matrix Transformations in Convolutional Neural Networks

Alberti, Michele; Botros, Angela; Schütz, Narayan; Ingold, Rolf; Liwicki, Marcus; Seuret, Mathias (5 May 2021). Trainable Spectrally Initializable Matrix Transformations in Convolutional Neural Networks. In: 2020 25th International Conference on Pattern Recognition (ICPR) (pp. 8204-8211). IEEE 10.1109/ICPR48806.2021.9412204

[img] Text
Alberti__2021_.pdf - Published Version
Restricted to registered users only
Available under License Publisher holds Copyright.

Download (2MB) | Request a copy

In this work, we introduce a new architectural component to Neural Network (NN), i.e., trainable and spectrally initializable matrix transformations on feature maps. While previous literature has already demonstrated the possibility of adding static spectral transformations as feature processors, our focus is on more general trainable transforms. We study the transforms in various architectural configurations on four datasets of different nature: from medical (ColorectalHist, HAM10000) and natural (Flowers) images to historical documents (CB55). With rigorous experiments that control for the number of parameters and randomness, we show that networks utilizing the introduced matrix transformations outperform vanilla neural networks. The observed accuracy increases appreciably across all datasets. In addition, we show that the benefit of spectral initialization leads to significantly faster convergence, as opposed to randomly initialized matrix transformations. The transformations are implemented as auto-differentiable PyTorch modules that can be incorporated into any neural network architecture. The entire code base is open-source.

Item Type:

Conference or Workshop Item (Paper)

Division/Institute:

10 Strategic Research Centers > ARTORG Center for Biomedical Engineering Research > ARTORG Center - Gerontechnology and Rehabilitation

Graduate School:

Graduate School for Cellular and Biomedical Sciences (GCB)

UniBE Contributor:

Botros, Angela Amira and Schütz, Narayan

Subjects:

000 Computer science, knowledge & systems
500 Science > 510 Mathematics

ISBN:

978-1-7281-8808-9

Publisher:

IEEE

Language:

English

Submitter:

Aileen Näf

Date Deposited:

25 Jun 2021 14:35

Last Modified:

25 Jun 2021 14:35

Publisher DOI:

10.1109/ICPR48806.2021.9412204

BORIS DOI:

10.48350/156852

URI:

https://boris.unibe.ch/id/eprint/156852

Actions (login required)

Edit item Edit item
Provide Feedback