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。

Combine statistical and symbolic artificial intelligence techniques

181 点作者 ghosthamlet大约 6 年前

11 条评论

mark_l_watson大约 6 年前
I work in the field of deep learning but in the 1980s and 1990s I used Common Lisp and worked on symbolic AI projects.<p>For several years, my gut instinct has been that the two technologies should be combined. Since neural nets are basically functions, I think it makes sense to compose functional programs using network models for perception, word and graph embedding, etc.<p>EDIT: I can’t wait to see the published results in May! EDIT 2: another commenter reelin posted a link to the draft paper <a href="https:&#x2F;&#x2F;openreview.net&#x2F;pdf?id=rJgMlhRctm" rel="nofollow">https:&#x2F;&#x2F;openreview.net&#x2F;pdf?id=rJgMlhRctm</a>
评论 #19583794 未加载
js8大约 6 年前
In my view, this is the endgame, really. Take any numerical technique, at the level of computers we always work with discrete bits. So you can reformulate any numerical problem (such as a problem of finding a probability distribution) on floats in terms of operations on individual bits, i.e. as a purely symbolic calculation.<p>However, doing so can very quickly lead to intractable problems of resolving satisfiability. So until we either manage to tame NP problems somehow (either by generating only easy instances, or by proving P=NP), we will always have to add some linearity assumptions (i.e. use numerical quantities) somewhere, and it will always be a bit of a mystery whether it actually helped to solve the problem or not.<p>In other words, we use statistics to overcome (inherent?) intractability, but in the process we add bias (as a trade-off). This is not necessarily bad, since it can help to actually solve a real problem. However, for any new problem, we will have understand the trade-offs again.
评论 #19581285 未加载
xvilka大约 6 年前
There is an interesting project - DeepProbLog[1], based on the ProbLog[2] (a Prolog dialect with probabilistic reasoning) and Deep Learning combined. I only wish it was Rust, so it would have been safer, faster, and easier to embed in your programs. I have high hopes to the Scryer Prolog[3], and it seems[4] the author think about probabilistic extensions too.<p>[1] <a href="https:&#x2F;&#x2F;bitbucket.org&#x2F;problog&#x2F;deepproblog" rel="nofollow">https:&#x2F;&#x2F;bitbucket.org&#x2F;problog&#x2F;deepproblog</a><p>[2] <a href="https:&#x2F;&#x2F;dtai.cs.kuleuven.be&#x2F;problog&#x2F;" rel="nofollow">https:&#x2F;&#x2F;dtai.cs.kuleuven.be&#x2F;problog&#x2F;</a><p>[3] <a href="https:&#x2F;&#x2F;github.com&#x2F;mthom&#x2F;scryer-prolog" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;mthom&#x2F;scryer-prolog</a><p>[4] <a href="https:&#x2F;&#x2F;github.com&#x2F;mthom&#x2F;scryer-prolog&#x2F;issues&#x2F;69" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;mthom&#x2F;scryer-prolog&#x2F;issues&#x2F;69</a>
评论 #19579518 未加载
chalst大约 6 年前
Excellent.<p>I have a general concern that some working with ML don&#x27;t appreciate the experience and technology that statisticians have developed to deal with bias, which I think is the biggest problem in the field. I tweeted &quot;ML is v impressive, but has no automated way to ensure no bias. Statistical modelling can&#x27;t match ML for parameter dimensions, but it can make explicit what is going on with the parameters you have and the assumptions you have. But advantages of theft over honest toil...&quot; - some of the responses in the thread are interesting.<p>My original tweet: <a href="https:&#x2F;&#x2F;twitter.com&#x2F;txtpf&#x2F;status&#x2F;1102437933301272577" rel="nofollow">https:&#x2F;&#x2F;twitter.com&#x2F;txtpf&#x2F;status&#x2F;1102437933301272577</a><p>Bob Watkins&#x27; tweet: <a href="https:&#x2F;&#x2F;twitter.com&#x2F;bobwatkins&#x2F;status&#x2F;1102568735485972480" rel="nofollow">https:&#x2F;&#x2F;twitter.com&#x2F;bobwatkins&#x2F;status&#x2F;1102568735485972480</a>
inetsee大约 6 年前
The questions about object relationships sound a lot like SHRDLU[1] which dates back about 50 years ago.<p>[1] <a href="https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;SHRDLU" rel="nofollow">https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;SHRDLU</a>
bglusman大约 6 年前
Reminds me of a recent comment I saw but can&#x27;t find by Douglas Lenat (of Cyc[1] fame, also relevant here) about how all the work on deep learning was great but now we need to marry the two, much like the ideas about how the &quot;right brain&quot; and &quot;left brain&quot; or system 1 and system 2 or something work together and work differently but we couldn&#x27;t very well function as humans without both.<p>[1]<a href="https:&#x2F;&#x2F;en.m.wikipedia.org&#x2F;wiki&#x2F;Cyc" rel="nofollow">https:&#x2F;&#x2F;en.m.wikipedia.org&#x2F;wiki&#x2F;Cyc</a>
taeric大约 6 年前
Soon we&#x27;ll be combining statistical, symbolic, and algorithmic intelligence techniques. I question why that isn&#x27;t the assumed position. :(<p>That is to say, we have devised some algorithms that are truly impressive. There is little reason to think an intelligence couldn&#x27;t devise them, of course. There is also little reason I can see, to not think we could help out programs by providing them.
评论 #19584426 未加载
评论 #19580929 未加载
nwatson大约 6 年前
Reminiscent of fuzzy logic: <a href="https:&#x2F;&#x2F;en.m.wikipedia.org&#x2F;wiki&#x2F;Fuzzy_logic" rel="nofollow">https:&#x2F;&#x2F;en.m.wikipedia.org&#x2F;wiki&#x2F;Fuzzy_logic</a><p>The Wikipedia article discusses various extensions of logic and symbolic computation to include probabilistic elements. This was a popular topic in the early 90s.
Reelin大约 6 年前
For anyone who&#x27;d prefer a direct link to the conference paper this seems to be based on: <a href="https:&#x2F;&#x2F;openreview.net&#x2F;forum?id=rJgMlhRctm" rel="nofollow">https:&#x2F;&#x2F;openreview.net&#x2F;forum?id=rJgMlhRctm</a>
评论 #19582083 未加载
JabavuAdams大约 6 年前
So, I&#x27;ve privately been working along similar lines, although I haven&#x27;t published anything, and I also haven&#x27;t read their specific approach.<p>How do I prevent a situation where I can&#x27;t work on my hobby project of multiple years because this stuff gets patented?
评论 #19584031 未加载
fizixer大约 6 年前
Yeah, feel free to dive into my past comments. I probably said many years ago that a combo of ML and GOFAI has massive potential, in a wide range of applications.
评论 #19582523 未加载