深度分享：世界頂級計算語言學科學家Ken Church CCL 2018主旨報告

知識 12-15

來源：TsinghuaNLP

本文多圖，建議閱讀10分鐘。

本文為你分享世界頂級計算語言學科學家在CCL2018的報告內容。

報告人：Kenneth Ward Church (Baidu Research Fellow)

題目：Minsky, Chomsky & Deep Nets

摘要：When Minsky and Chomsky were at Harvard in the 1950s, they started out their careers questioning a number of machine learning methods that have since regained popularity. Their objections are remembered as too negative: ngrams can"t do this and nets can"t do that. But more constructively, their arguments led to what is now known as the Chomsky Hierarchy and time and space complexity. The deep nets literature needs a high-level organization like the Chomsky Hierarchy. There are so many results being published these days that it is hard to see the big picture.My colleagues at Baidu have proposed a metric for ranking deep net problems by a metric of difficulty. There is no data like more data, but more data helps more for some problems than others. Their ranking suggests that more data is particularly helpful for problems that people are really excited about in speech and vision, and less so for problems in computational linguistics.

https://stu.cs.tsinghua.edu.cn/wiki/images/4/45/2018-10-21_CCL2018_Ken.pdf

PPT 內容如下