Neural Turing Machines [pdf] | Hacker News Reader

Neural Turing Machines [pdf]

arxiv.org

75 pointsneurologic11 comments

For arXiv submissions, please submit the abstract link; in this case: http://arxiv.org/abs/1410.5401

The advantage is immediate: One gets a plain website with the abstract, which allows the visitor to decide whether the PDF download is worth it. The PDF itself is only one more click away.

I personally would also leave out the v1, as that link automatically goes to the latest version in all cases.

I strongly prefer the direct link to the paper. Downloading and viewing a PDF is not so burdensome that this is a problem. I get the abstract at the top of the paper and if I'm no interested I can just close the tab that's showing the PDF same as closing a web page.

I strongly prefer the cite-able link to the pdf. The link points to the PDF, but the PDF does not point to the link. The link points to all versions of the paper.

As it turns out, the URL of the pdf is cite-able.

But it doesn't link to the versions of the paper.

You want to save one click of the mouse. I want to save having to reverse-engineer the parent URL.

Serious question: are there any browsers in common use that don't display PDFs inline? Chrome and Firefox both do, and their built-in PDF viewers seem very simple and streamlined and efficient.

It just seems a bit odd to question whether a 760KB PDF is "worth it" when the current top post on HN is a blog post that downloads 780KB of Javascript, and has attracted no such comments.

Anyway, the paper is very interesting. The sorting example is particularly impressive -- I almost wouldn't have believed it was possible to learn automatically.

The browser on my phone and tablet

I'm running Firefox on a fairly beefy laptop, and the integrated PDF viewer annoyed me so much that I disabled it some time last year. I suppose it's fine for simple stuff, and it's definitely better than an Adobe Reader plugin, but it just manages to be much too sluggish on too many real-world examples out there when standalone viewers work just fine and smooth - but unfortunately don't integrate well in the overall browsing experience.

I'm a serial computer and browser abuser, and on OS X, I can easily crash my browser, and sometimes even the whole OS, if I'm not careful what I load up in a new browser tab. So I'd rather have the simpler abstract page too.

Interesting read. Extrapolating from how the network learned to copy and sort data it points a bit to a future with software engineers not coding programs to do the job but training networks with the right data instead. That would basically end the language war. :) I hope time to silicon is a single digit year.

Wow, if they are publishing this, you can bet they are applying these techniques to much more sophisticated problems already.

This might be another preliminary signal for a future wave of white-collar job destruction.

Reply on news.ycombinator.com