colah3

13 ms·

91.
by colah3 10y ago
That's a nice connection -- it does have a similar flavor to that kind of humanities scholarship. :)
92.
by colah3 10y ago
> 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
93.
by colah3 10y ago
Karoly does lovely work! :)
94.
by colah3 10y ago
Our present JIF is undefined because we haven't existed for two years yet. If you just apply the formulas anyways, you'll get an JIF of (6 citations)/(4 publications) = 1.5. Again, this number is really pessimistic because th
95.
by colah3 10y ago
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/augment
96.
by colah3 10y ago
Every article has a DOI: https://search.crossref.org/?q=Distill&publication=Distill
97.
by colah3 10y ago
We're interested in both review/tutorial articles and novel research articles. :)
98.
by colah3 10y ago
I think that having a venue that can publish non-traditional academic artifacts is an important step for reproducibility, even if it isn't our focus. It seems clear to me that the future will involve some kind of linking reproducibilit
99.
by colah3 10y ago
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
100.
by colah3 10y ago
Yes. Everything is published under Creative Commons Attribution. (One of the members of our steering committee, Michael Nielsen, has a significant history advocating for open science. I think there's about a snowball's chance in h
101.
by colah3 10y ago
Thank you. Talking to others in the space, I think a number of people share my experience, and often stopped doing this kind of work. While there's extremely positive feedback from some people, it's kind of "you have a cool h
102.
by colah3 10y ago
All we can do is work hard to build academic support: * In the last three weeks, we've had 80 outreach conversations with various stakeholders for Distill. The majority of these have been academic researchers. The response has been ext
103.
by colah3 10y ago
Various announcements: Google Research: https://research.googleblog.com/2017/03/distill-supporting-c... DeepMind: https://deepmind.com/blog/distill-communicating-science-mach... OpenAI: http
104.
by colah3 10y ago
The "deliberateness" or "conciencousness" of a lock is an interesting dimension. A laser cutter I used to work with had two switches, one of which was a safety switch (like [1]) to prevent accidental activation. The goal
105.
by colah3 10y ago
That's a really nice distinction. It feels like there's a bit of a spectrum, though. The lock on a houses door won't prevent a determined criminal, but it sends a very strong signal and is inconvenient to break. Perhaps we sh
106.
by colah3 10y ago
Some do, although this may be limited to high-end models? I have a friend who uses hearing aids. They can tune out background noise, focus on directions, turn off hearing, and even listen to magnetic fields (while debugging circuits). I&#x2
107.
by colah3 10y ago
Note that the contribution of this article is techniques for visualizing models, not the novel itself.
108.
by colah3 10y ago
Latin character only. It was trained on the IAM Handwriting database [1], which is all English handwriting. [1] http://www.fki.inf.unibe.ch/databases/iam-handwriting-databa...
109.
by colah3 10y ago
The generative model of handwriting that we're working with here probably isn't very applicable to handwriting recognition. In principle, you could train a model like this to jointly model the text and the produced handwriting and
110.
by colah3 10y ago
One thing that's interesting about this article is that it's the canonical publication for a novel result. It isn't a blog post accompanying a paper -- it is the paper.
111.
by colah3 10y ago
t-SNE is used very widely in ML, both to analyze input data and to analyze learned representations. To me, the canonical example of people using t-SNE this way is visualizing word embeddings, like this: http://metaoptimize.s3.ama
112.
by colah3 10y ago
Paper: https://arxiv.org/pdf/1606.06565v1.pdf Google Post: https://research.googleblog.com/2016/06/bringing-precision-t... It was a pleasure for us to work on this with OpenAI and others. Joh
113.
by colah3 10y ago
I hope my previous comment didn't come across as me suggesting that the paper shouldn't be scrutinized because it comes from an established research group. Of course, it absolutely should be and double blind review is a helpful me
114.
by colah3 10y ago
I missed the actual paper before it got taken down. But as someone in the field, I'd like to say that these are real researchers at a respected group. The Helsinki group has done a lot of lovely work over the last few years -- ladder n
115.
by colah3 10y ago
I'm not the GP, but I have a similar impression that GiveDirectly has really strong evidence supporting it. GiveWell -- a separate organization that does in depth analysis of giving opportunities -- has a detailed write up on them: ht
116.
by colah3 10y ago
Although, there was at one point an anarchist organization called the weather underground: https://en.wikipedia.org/wiki/Weather_Underground I understand they had a "brick reallocation committee" that realloc
117.
by colah3 10y ago
I drew the diagrams in Inkscape. (Naturally, this doesn't include the screenshots of TensorBoard or EEG.)
118.
by colah3 11y ago
I really like org-mode and used it heavily for about a year. But I found having my todo list on a single computer pretty limiting. Thankfully, there's a new generation of org-mode inspired cloud apps: * https://workflowy.com
119.
by colah3 11y ago
Breath taking. This technique could be used to visually prove many interesting properties of the continuous fourier transform. For example, you can visually prove that the fourier transform is its own inverse, up to a constant factor. It&#x
120.
by colah3 11y ago
As an industry outsider, it wasn't obvious to me that their business plan was viable, even if the technology panned out. They want to charge $5,000 for a round trip between NYC and London. That's at least 5x the cost of getting a

More ›