C. Etmann. (2019)cite arxiv:1906.06637Comment: 16 pages, 7 figures.
Аннотация
In recent years, an increasing number of neural network models have included
derivatives with respect to inputs in their loss functions, resulting in
so-called double backpropagation for first-order optimization. However, so far
no general description of the involved derivatives exists. Here, we cover a
wide array of special cases in a very general Hilbert space framework, which
allows us to provide optimized backpropagation rules for many real-world
scenarios. This includes the reduction of calculations for
Frobenius-norm-penalties on Jacobians by roughly a third for locally linear
activation functions. Furthermore, we provide a description of the
discontinuous loss surface of ReLU networks both in the inputs and the
parameters and demonstrate why the discontinuities do not pose a big problem in
reality.
Описание
A Closer Look at Double Backpropagation - 1906.06637.pdf
%0 Generic
%1 etmann2019closer
%A Etmann, Christian
%D 2019
%K double-backpropagation from:adulny maths theory
%T A Closer Look at Double Backpropagation
%U http://arxiv.org/abs/1906.06637
%X In recent years, an increasing number of neural network models have included
derivatives with respect to inputs in their loss functions, resulting in
so-called double backpropagation for first-order optimization. However, so far
no general description of the involved derivatives exists. Here, we cover a
wide array of special cases in a very general Hilbert space framework, which
allows us to provide optimized backpropagation rules for many real-world
scenarios. This includes the reduction of calculations for
Frobenius-norm-penalties on Jacobians by roughly a third for locally linear
activation functions. Furthermore, we provide a description of the
discontinuous loss surface of ReLU networks both in the inputs and the
parameters and demonstrate why the discontinuities do not pose a big problem in
reality.
@misc{etmann2019closer,
abstract = {In recent years, an increasing number of neural network models have included
derivatives with respect to inputs in their loss functions, resulting in
so-called double backpropagation for first-order optimization. However, so far
no general description of the involved derivatives exists. Here, we cover a
wide array of special cases in a very general Hilbert space framework, which
allows us to provide optimized backpropagation rules for many real-world
scenarios. This includes the reduction of calculations for
Frobenius-norm-penalties on Jacobians by roughly a third for locally linear
activation functions. Furthermore, we provide a description of the
discontinuous loss surface of ReLU networks both in the inputs and the
parameters and demonstrate why the discontinuities do not pose a big problem in
reality.},
added-at = {2021-02-26T16:08:31.000+0100},
author = {Etmann, Christian},
biburl = {https://www.bibsonomy.org/bibtex/266c6dd1ffb001a1514a76d66a8fb1a1c/adulny},
description = {A Closer Look at Double Backpropagation - 1906.06637.pdf},
interhash = {f5629b1d234299d3895c2b35c1e598b1},
intrahash = {66c6dd1ffb001a1514a76d66a8fb1a1c},
keywords = {double-backpropagation from:adulny maths theory},
note = {cite arxiv:1906.06637Comment: 16 pages, 7 figures},
timestamp = {2021-02-26T16:08:31.000+0100},
title = {A Closer Look at Double Backpropagation},
url = {http://arxiv.org/abs/1906.06637},
year = 2019
}