[2303.11835] Lipschitz-bounded 1D convolutional neural networks utilizing the Cayley rework and the controllability Gramian

0
41


Obtain a PDF of the paper titled Lipschitz-bounded 1D convolutional neural networks utilizing the Cayley rework and the controllability Gramian, by Patricia Pauli and three different authors

Obtain PDF

Summary:We set up a layer-wise parameterization for 1D convolutional neural networks (CNNs) with built-in end-to-end robustness ensures. In doing so, we use the Lipschitz fixed of the input-output mapping characterised by a CNN as a robustness measure. We base our parameterization on the Cayley rework that parameterizes orthogonal matrices and the controllability Gramian of the state area illustration of the convolutional layers. The proposed parameterization by design fulfills linear matrix inequalities which might be ample for Lipschitz continuity of the CNN, which additional permits unconstrained coaching of Lipschitz-bounded 1D CNNs. Lastly, we practice Lipschitz-bounded 1D CNNs for the classification of coronary heart arrythmia information and present their improved robustness.

Submission historical past

From: Patricia Pauli [view email]
[v1]
Mon, 20 Mar 2023 12:25:43 UTC (185 KB)
[v2]
Thu, 25 Jan 2024 09:35:25 UTC (82 KB)



Supply hyperlink

LEAVE A REPLY

Please enter your comment!
Please enter your name here