Profile cover photo
Profile photo
Darren Cook
90 followers -
I do interesting things with data...
I do interesting things with data...

90 followers
About
Posts

Post has attachment

Post has attachment
I've been working hard on this new book for the past 2-3 months, so it is about time I announced it properly!

It is another book for O'Reilly, and hard-core data science and AI this time. Well "hard-core" but with none of the maths, only practical examples and advice. The H2O refers to http://h2o.ai/, which is a really nice Big Data machine learning platform: powerful, easy-to-use, and really growing in popularity.
(The cover animal is a crayfish; publication in a couple of months, or so. I will have discount codes! Ask me!)
Photo

Anyone I know going to London Strata, June 1/2/3 ? (It is a Big Data/Hadoop/etc. conference.)

I'll be there "hosting" a room on the Friday. But also planning to be there on the Thursday, and hopefully on the Wednesday too.

Post has attachment
Matt Dowle was testing how his merge algorithm scales, merging 10 billion rows with 10 billion rows (!!), in an H2O cluster. But it was taking 20 minutes, when 10 minutes felt more reasonable to him. Lots of algorithm analysis followed:
    http://blog.h2o.ai/2016/05/red-herring-bites

... it turned out to be node 1 in the 10-node cluster was running slower: it was running gigabit ethernet, all the rest were running 10G-ethernet. It reminded me I had had the same problem with slow file copy, many years ago, but at a different scale: a 100M connection turned out to be running at 10M. I swapped the cable and everything suddenly got quicker!

BTW, I'm doing a lot with H2O (scalable machine learning) recently, and hoping to have an exciting announcement on that topic very soon. Watch This Space.

Post has attachment

Post has attachment
In this blog post I introduced and compared three libraries for working with WebGL, that I have experience using in real-world projects: Three.JS, Babylon.JS and Superpowers:

http://darrendev.blogspot.co.uk/2016/03/comparison-of-three-webgl-libraries.html

Post has attachment
A very different blog topic from the previous one (which was on high-precision arithmetic in the R language); this one shows how to make gradients on 3D objects in the browser (using the WebGL library called Three.JS):

http://darrendev.blogspot.co.uk/2016/03/gradients-in-threejs.html

It is a lovely effect, but also quite a lot of code to achieve it, so if you know a better way, let me know (here, or in the comments on the Darren's Developer Diary blog).

Post has attachment
I've a big backlog of technical blog posts to publish; here is the first, on how to use very large numbers, and fractions (aka rationals) in the R language.

Post has attachment
At last, someone is getting some proper publicity for the human-rights abuse that is the current UK immigration policy of splitting up families.

If you cannot contribute to the film production costs, please do share this with other people, to help raise some awareness. I'm finding there are many UK residents who are not aware of this rule: I need to explain it twice, because they think what I've just explained cannot possibly be the way it works.

(I'm sure you all already know the stress this policy and lost work-time has caused me over the past year, not to mention the huge financial cost - we had to sell our house in Tokyo, which ended up having to be in an "accept any offer" state, to meet the financial requirements.)

Post has attachment
Don't get stuck on the headline here; the comments quickly debunk it. My favourite comment was: "So yes, if you do stupid things, you can make bad engineering decisions look like good ones. "

A lot of governmental economic and military policy is like that, around the world. It is a shame that those policies don't get to be peer-reviewed by intelligent people, who have no vested interest, before being implemented!

http://developers.slashdot.org/story/15/03/25/1430251/no-its-not-always-quicker-to-do-things-in-memory
Wait while more posts are being loaded