When is memorization of irrelevant training data necessary for high-accuracy learning?
When is memorization of irrelevant training data necessary for high-accuracy learning?
复制标题
什么时候为了高精度学习需要记忆不相关的训练数据?
DOI:
10.1145/3406325.3451131
复制
发表时间:
2021
期刊:
影响因子:
--
通讯作者:
Talwar, Kunal
中科院分区:
文献类型:
--
作者:
Brown, Gavin;Bun, Mark;Feldman, Vitaly;Smith, Adam;Talwar, Kunal
Modern machine learning models are complex and frequently encode surprising amounts of information about individual inputs. In extreme cases, complex models appear to memorize entire input examples, including seemingly irrelevant information (social security numbers from text, for example). In this paper, we aim to understand whether this sort of memorization is necessary for accurate learning. We describe natural prediction problems in which every sufficiently accurate training algorithm must encode, in the prediction model, essentially all the information about a large subset of its training examples. This remains true even when the examples are high-dimensional and have entropy much higher than the sample size, and even when most of that information is ultimately irrelevant to the task at hand. Further, our results do not depend on the training algorithm or the class of models used for learning.Our problems are simple and fairly natural variants of the next-symbol prediction and the cluster labeling tasks. These tasks can be seen as abstractions of text- and image-related prediction problems. To establish our results, we reduce from a family of one-way communication problems for which we prove new information complexity lower bounds.
登录
查看更多内容
DOI:
--
发表时间:
2014
期刊:
SIAM journal on computing (Print)
影响因子:
--
作者:
V. Feldman;David Xiao
通讯作者:
David Xiao
DOI:
--
发表时间:
2020
期刊:
Neural Information Processing Systems
影响因子:
--
作者:
Roi Livni;S. Moran
通讯作者:
S. Moran
DOI:
--
发表时间:
2019-02
期刊:
ArXiv
影响因子:
--
作者:
Chiyuan Zhang;Samy Bengio;Moritz Hardt;Y. Singer
通讯作者:
Chiyuan Zhang;Samy Bengio;Moritz Hardt;Y. Singer
影响因子:
6
作者:
A. Beimel;Kobbi Nissim;Uri Stemmer
通讯作者:
Uri Stemmer
影响因子:
1.2
作者:
J. Littlewood
通讯作者:
J. Littlewood