In recent years, Deep Learning has become the go-to solution for a broad
range of applications, often outperforming state-of-the-art. However, it is
important, for both theoreticians and practitioners, to gain a deeper
understanding of the difficulties and limitations associated with common
approaches and algorithms. We describe four families of problems for which some
of the commonly used existing algorithms fail or suffer significant difficulty.
We illustrate the failures through practical experiments, and provide
theoretical insights explaining their source, and how they might be remedied.
%0 Generic
%1 shalevshwartz2017failures
%A Shalev-Shwartz, Shai
%A Shamir, Ohad
%A Shammah, Shaked
%D 2017
%K deep learning limitations
%T Failures of Deep Learning
%U http://arxiv.org/abs/1703.07950
%X In recent years, Deep Learning has become the go-to solution for a broad
range of applications, often outperforming state-of-the-art. However, it is
important, for both theoreticians and practitioners, to gain a deeper
understanding of the difficulties and limitations associated with common
approaches and algorithms. We describe four families of problems for which some
of the commonly used existing algorithms fail or suffer significant difficulty.
We illustrate the failures through practical experiments, and provide
theoretical insights explaining their source, and how they might be remedied.
@misc{shalevshwartz2017failures,
abstract = {In recent years, Deep Learning has become the go-to solution for a broad
range of applications, often outperforming state-of-the-art. However, it is
important, for both theoreticians and practitioners, to gain a deeper
understanding of the difficulties and limitations associated with common
approaches and algorithms. We describe four families of problems for which some
of the commonly used existing algorithms fail or suffer significant difficulty.
We illustrate the failures through practical experiments, and provide
theoretical insights explaining their source, and how they might be remedied.},
added-at = {2017-03-24T20:19:43.000+0100},
author = {Shalev-Shwartz, Shai and Shamir, Ohad and Shammah, Shaked},
biburl = {https://www.bibsonomy.org/bibtex/287497d3648b14dd62091b20abaa82cd2/tomhanika},
description = {Failures of Deep Learning},
interhash = {22c00ab689adda4ae2b21252b9d0998b},
intrahash = {87497d3648b14dd62091b20abaa82cd2},
keywords = {deep learning limitations},
note = {cite arxiv:1703.07950},
timestamp = {2017-03-24T20:19:43.000+0100},
title = {Failures of Deep Learning},
url = {http://arxiv.org/abs/1703.07950},
year = 2017
}