Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Distill: a modern machine learning journal

distill.pub

Post reply on HN
Print view View original

81–90 of 109 posts

Re: Distill: a modern machine learning journal

#81

Post by Naracion »

colah3 wrote: post #22
j2kun wrote: post #18

I sure hope this catches on, but we should all be aware of the hurdles: - Little incentive for researchers to do this beyond their own good will. - Most ML researchers are bad writers, and it's unlikely that the editing team will do the work needed (which is often a larger reorganization of a paper and ideas) to improve clarity. - Producing great writing and clear, interactive figures, and managing an ongoing github…

You're absolutely right that this is a lot of work, and not many ML researchers have all the skills needed for it. In the short term, Distill's editorial assistance will help authors produce outstanding papers, although they need to be willing to work as well. In the longer-term, I'd like to explore match making between data visualization people who would like to get into machine learning and machine learning researc…

As another designer + researcher with a varied background and an interest in data viz as well as ML, I am super interested in this as a potential contributor. I have experience creating an interactive visualization interface for simple ML algorithms (which has been used by professors in the life sciences department to understand / get a new perspective on what's happening). I would LOVE to be able to be involved with Distill.

I have actually been meaning to write a paper on my findings and have been looking for journals to write for. However it doesn't quite "fit" with most journals. Distill looks like it's more catered to "professional" machine learning people, at least for now. Is there any way that somebody with my background (design+data viz+development+interest and curiosity to learn ML) could be involved with Distill?

Re: Distill: a modern machine learning journal

#82

Post by sytelus »

This is great but it would have been even better if Distill was designed to play well with the current system. Vast majority of researchers are focused on publishing at various conferences with strict deadlines. Even if they had all the skillsets and time to produce these beautiful illustrations, I highly doubt this will change.

Also, it is very likely that veterans in the field might think of this format as too verbose and too sugar coated, more appropriate for less math-savvy users and therefore not mainstream. Furthermore, I really feel TeX is irreplaceable unless you got all of its feature covered. All of the historic effort to replace TeX - even with bells and whistles of WYSIWYG editors - in research has failed and its important to learn from those failures. You will be surprised how many researchers insist on printing out the paper for reading even when they have access to tablets and PC.

Instead of being another peer reviewed journal, Distill could act as the following:

- platform to publish supplemental material and code

- platform to manage communication/issues post publication

- platform for readers to invite other readers for peer review and generate "front page" based on some sort of reviewer trust relationship.

- platform to host Python and MatLab code with web frontends without researchers having to learn new developer skills

- support pdf submissions but without all the eliteness of arxiv and using algorithms to create the "front page" based on some sort of peer reviewer rankings.

Above features are indeed sorely missing and Distill has good opportunity to become an "add-on" to current academic publishing systems as opposed to another peer reviewed journal.

Re: Distill: a modern machine learning journal

#83

Post by colah3 »

Naracion wrote: post #81
colah3 wrote: post #22

Earlier quoted context omitted.

You're absolutely right that this is a lot of work, and not many ML researchers have all the skills needed for it. In the short term, Distill's editorial assistance will help authors produce outstanding papers, although they need to be willing to work as well. In the longer-term, I'd like to explore match making between data visualization people who would like to get into machine learning and machine learning researc…

As another designer + researcher with a varied background and an interest in data viz as well as ML, I am super interested in this as a potential contributor. I have experience creating an interactive visualization interface for simple ML algorithms (which has been used by professors in the life sciences department to understand / get a new perspective on what's happening). I would LOVE to be able to be involved with…

> Is there any way that somebody with my background (design+data viz+development+interest and curiosity to learn ML) could be involved with Distill?

Absolutely. We know a number of leading ML researchers who would love to publish papers as Distill articles but don't have the design/data vis skills. We'd like to facilitate collaborations which would lead to data vis people co-authoring cutting edge research papers.

Re: Distill: a modern machine learning journal

#84

Post by legel »

KCFforecast wrote: post #47
j2kun wrote: post #18

I sure hope this catches on, but we should all be aware of the hurdles: - Little incentive for researchers to do this beyond their own good will. - Most ML researchers are bad writers, and it's unlikely that the editing team will do the work needed (which is often a larger reorganization of a paper and ideas) to improve clarity. - Producing great writing and clear, interactive figures, and managing an ongoing github…

I think you have emphasized the main point: a lot of work for a low reward. Research is more above exploring the state of the art and new venues, divulgation and graphics is more akin to book sellers (for example Nielsen open science, and other interesting books, but for young researcher the most important and rewarding goal is to publish.

I think it depends on what type of researcher you are. In every field there are always authoritative leaders who are comfortable writing "survey papers", which is perhaps most comparable to what the "research distiller" is all about. Except, these guys know from experience that visualization of complexity is perhaps the most direct way of communicating to the brain... and the real-time interactive nature of such technologies is far beyond "book sellers", and more into how you can imagine the future of human communication more generally approaching (perhaps with support of real-time speech recognition and graphics generation AI, e.g.)... but I digress - this is most certainly a fantastic move in the right direction for the research community at large, and especially for the machine learning community where so much is happening so fast, and we really do need people to stop and help us "distill". :) I have fond memories of finally understanding LSTMs based on Christopher Olah's blog, and if we can somehow scale this up and out in other areas, I'll gladly invest time and money and energy into helping pursue the bigger opportunities here...

Re: Distill: a modern machine learning journal

#85

Post by j2kun »

MLasstProf wrote: post #79
j2kun wrote: post #18

I sure hope this catches on, but we should all be aware of the hurdles: - Little incentive for researchers to do this beyond their own good will. - Most ML researchers are bad writers, and it's unlikely that the editing team will do the work needed (which is often a larger reorganization of a paper and ideas) to improve clarity. - Producing great writing and clear, interactive figures, and managing an ongoing github…

Thanks for bringing these points up j2kun. I'm a junior faculty working in ML with no personal knowledge of web development, d3, etc. While the papers currently on Distill are absolutely gorgeous and will be an invaluable tool for learning advanced ML concepts, I simply cannot see myself or my students putting the time to actually create something like that. Unless a student is especially adept at the specific tools…

From my experience, most ML researchers are in your camp. They are primarily interested in the ML, and good (not-just-in-your-head) visualizations are at best icing on the cake of their understanding.

Re: Distill: a modern machine learning journal

#86

Post by j2kun »

eshvk wrote: post #74
FractalNerve wrote: post #71

Earlier quoted context omitted.

Well, now you need a Distill WYSIWYG, to make it usable (for most of the intended audience). Hey let's be honest, most academics (that I know) still don't even use LaTeX (or refuse to do so). This is really cool, but requires way too many skills (in js/css3/html5/distill-extensions and node.js). Personally, my team and I had really great experience with sharelatex.com, whom only I had knowledge about LaTeX. I liked t…

> Hey let's be honest, most academics (that I know) still don't even use LaTeX (or refuse to do so). What field? TeX is pretty much de rigueur in Math/CS/Physics graduate schools in the U.S.

I agree with you here, and have no idea what the OP is talking about. In Math and ML, TeX is so ingrained in the culture that there are _jokes_ based on TeX puns. (Where do mathematicians go for a rational rack of ribs? The \mathbb Q)

Re: Distill: a modern machine learning journal

#87

Post by Blahah »

colah3 wrote: post #77
timClicks wrote: post #76

Earlier quoted context omitted.

I feel like binding the journal to GitHub means that it's less likely to exist over the long term (where long term means >100 years, which is as long as I would expect an academic article to be accessible for).

We produce "archive html" files where everything is bundled into a single file. We're looking into ensuring their long-term preservation with projects like LOCKSS. Example: http://distill.pub/2016/augmented-rnns/index.archive.html

A simple intermediate step would be to archive with Zenodo

Re: Distill: a modern machine learning journal

#88

Post by djabatt »

It would be cool to see greater diversity of thinking on the about page. perhaps the pub is designed for insiders.

Having more research transparency is great for community of likes minds to learn from. A suggested addition is an section and team to lead a discussion ML ethics.

Re: Distill: a modern machine learning journal

#89

Post by fwx »

This is amazing! My burning question - as has been pointed out in the thread, the effort to produce a great article on Distill - generating interactive figures, doing front end web dev etc. would require a lot of time and resources on the part of the researchers. Is it possible to include within Distill an option to connect researchers to willing-and-able developers in those domains (for example, me) to help them get it done?

Re: Distill: a modern machine learning journal

#90

Post by kriro »

eshvk wrote: post #74
FractalNerve wrote: post #71

Earlier quoted context omitted.

Well, now you need a Distill WYSIWYG, to make it usable (for most of the intended audience). Hey let's be honest, most academics (that I know) still don't even use LaTeX (or refuse to do so). This is really cool, but requires way too many skills (in js/css3/html5/distill-extensions and node.js). Personally, my team and I had really great experience with sharelatex.com, whom only I had knowledge about LaTeX. I liked t…

> Hey let's be honest, most academics (that I know) still don't even use LaTeX (or refuse to do so). What field? TeX is pretty much de rigueur in Math/CS/Physics graduate schools in the U.S.

To my surprise certain subfields of CS don't use LaTeX at all or rarely and use MS-Word instead. You kind of have no choice since the conference/journal templates are only provided in one format (well you can create your own template but only if they accept PDF entries...yes some only accept Word files).
Post reply on HN