TE
科技回声
首页24小时热榜最新最佳问答展示工作
GitHubTwitter
首页

科技回声

基于 Next.js 构建的科技新闻平台,提供全球科技新闻和讨论内容。

GitHubTwitter

首页

首页最新最佳问答展示工作

资源链接

HackerNews API原版 HackerNewsNext.js

© 2025 科技回声. 版权所有。

Neurosymbolic AI Advances State of the Art on Math Word Problems

67 点作者 daram超过 5 年前

4 条评论

icandoit超过 5 年前
Searching for Nuerosymbolic I can only find this paper from 2017 at Microsoft Reasearch. It is from 2017 and uses an RNN to incrementally grows some input source code.<p><a href="https:&#x2F;&#x2F;www.microsoft.com&#x2F;en-us&#x2F;research&#x2F;publication&#x2F;neuro-symbolic-program-synthesis-2&#x2F;" rel="nofollow">https:&#x2F;&#x2F;www.microsoft.com&#x2F;en-us&#x2F;research&#x2F;publication&#x2F;neuro-s...</a><p>&gt;The first module, called the cross correlation I&#x2F;O network, given a set of input-output examples, produces a continuous representation of the set of I&#x2F;O examples. The second module, the RecursiveReverse-Recursive Neural Network (R3NN), given the continuous representation of the examples, synthesizes a program by incrementally expanding partial programs<p>I think stuff like this is very interesting and look forward to a future of old-hats complaining that new programmers let the intellisense write all their code for them.
deep_etcetera超过 5 年前
This article seems pointless without detailing any of the results, why not give some numbers on their performance on benchmarks?
leereeves超过 5 年前
Is the peer-reviewed paper available anywhere yet?<p>This article doesn&#x27;t tell us anything about the Neurosymbolic AI approach except that it performs well and &quot;essentially combines two existing techniques: neural attention Transformers...and tensor product representation&quot;.
评论 #21722078 未加载
评论 #21722842 未加载
jdmccaffrey超过 5 年前
(I contributed to the PureAI article) The research article is still under peer review so that&#x27;s why no detailed accuracy information or explanation of the technique can be published yet. If the PureAI article gave details, there&#x27;d be nothing to prevent an unscrupulous person from hijacking the content and establishing copyright&#x2F;patent for the intellectual property. I have seen the research paper and it presents full information about the technique and the experimental results.<p>James M