Representer Theorems for the Design of Deep Neural Networks and the Resolution of Continuous-Domain Inverse Problems
M. Unser
Ninth International Conference on Machine Learning in Medical Imaging (MLMI'18), Granada, Kingdom of Spain, September 16, 2018, in press.
Please do not bookmark the In Press papers as content and presentation may differ from the published version.
In a few seconds, you should be redirected to the published version. The preprint version is still available here
Our purpose in this talk is to reenforce the (deep) connection between splines and learning techniques. To that end, we first describe a recent representer theorem that states that the extremal points of a broad class of linear inverse problems with generalized total-variation regularization are adaptive splines whose type is linked to the underlying regularization operator L. For instance, when L is n-th derivative (resp., Laplacian) operator, the optimal reconstruction is a non-uniform polynomial (resp., polyharmonic) spline with the smallest possible number of adaptive knots.
The crucial observation is that such continuous-domain solutions are intrinsically sparse, and hence compatible with the kind of formulation (and algorithms) used in compressed sensing.
We then make the link with current learning techniques by applying the theorem to optimize the shape of individual activations in a deep neural network. By selecting the regularization functional to be the 2nd-order total variation, we obtain an "optimal" deep-spline network whose activations are piece-linear splines with a few adaptive knots. Since each spline knot can be encoded with a ReLU unit, this provides a variational justification of the popular ReLU architecture. It also suggests some new computational challenges for the determination of the optimal activations involving linear combinations of ReLUs.
References
-
M. Unser, J. Fageot, J.P. Ward, "Splines Are Universal Solutions of Linear Inverse Problems with Generalized TV Regularization," SIAM Review, vol. 59, no. 4, pp. 769-793, December 2017.
-
M. Unser, "A Representer Theorem for Deep Neural Networks," arXiv:1802.09210 [stat.ML].