Loading...
Loading...

Go to the content (press return)

Geometric bio-inspired deep learning

Author
Xambo, S.; Moya, E.
Type of activity
Presentation of work at congresses
Name of edition
9th International Congress on Industrial and Applied Mathematics
Date of publication
2019
Presentation's date
2020-07-16
Book of congress proceedings
ICIAM-2019: International Congress on Industrial and Applied Mathematics: Valencia, Spain: july 15-19: actas
First page
1
Last page
67
Abstract
Although Convolutional Neural Networks (CNN) are bio-inspired by neuroscientific principles, and that they can capture some essential properties of primary visual cortex V1, they still fall quite short when compared to the mammalians visual system, as the latter are resilient in face of more equivariant transformations, such as local invariant (equivariant) response to some changes in lighting and rotations. One possible reason is that the CNN the representation spaces in the current NN are usua...

Participants