In 1952 Alan Turing, a british mathematician, logician, cryptanalyst, and computer scientist, wrote a paper which remains influential in computational biology today. He explained how stripes might form on a snake’s skin [and other patterns on animals], using the dispersion of two chemicals; an activator [red] and an inhibitor [yellow]. The activator causes the colouration, and the inhibitor inhibits it. Turing wrote a pair of equations which say that concentrations of the activator cause creation of more inhibitor, but that the inhibitor diffuses and spreads out more quickly than the activator. As shown in the animation, this causes the activator to form peaks with surrounding basins of inhibitor. The concentrations of the two chemicals quickly converge to a stripey pattern where the red activator is periodically in higher concentration than the yellow inhibitor. [video] [more] [code]
The big surprise is that Google still uses the manually-crafted formula for its search results. They haven’t cut over to the machine learned model yet. Peter suggests two reasons for this. The first is hubris: the human experts who created the algorithm believe they can do better than a machine-learned model. The second reason is more interesting. Google’s search team worries that machine-learned models may be susceptible to catastrophic errors on searches that look very different from the training data. They believe the manually crafted model is less susceptible to such catastrophic errors on unforeseen query types.
- Anand Rajaraman in his blog post titled “Are Machine-Learned Models Prone to Catastrophic Errors?” (via foldmap)
My posthumus tumblr contribution, memkeys, is a top C++ project on github. Awesome!
Graceland in Memphis, TN. Blake was so excited.
Alan all settled in at Cuddles & Tails. Krissy is hiding somewhere.
Alan is doing just fine in some temp housing. In other news, we’ll use this blog to track our NYC to SF journey.
One of the few holdovers from my experience as an undergraduate university student is my ACM & IEEE memberships. I’ve maintained these memberships for years now without ever putting much thought into why I maintain them. A couple of weeks ago I mentioned on Twitter that I had been named a Senior Member of the Association of Computing Machinery (ACM). A few people followed up and basically were curious as to why I still maintain my membership, while others weren’t even really sure what these organizations were. If you didn’t spend some time in college, or in a more research focused organization, it’s unlikely that you have run into ACM/IEEE members on a regular basis. I think this is a shame.
While the ACM and IEEE both list a set of proposed benefits to membership on their respective sites, I think they both manage to miss the mark. I maintain my memberships for two reasons; opportunities for collaboration, and access to research. These also happen to be the two reasons why I most enjoyed being a university student; ample opportunities for collaboration with smart people, access to cutting edge research.
I’ve met a number of people through my membership that are either in some dark cave in academia, or buried in a bunker somewhere doing research. I’ve managed to be a part of some incredibly productive and positive joint work that otherwise might not have been possible outside of academia. This collaboration has opened my eyes to the theoretical, but my day job keeps me grounded in the practical. I find this gives me an advantage in my industry. This is valuable to me.
Additionally, the ACM/IEEE sponsor a huge number of conferences and journals. A lot of what is published today is targeted towards acceptance at these conferences and in these journals. My membership provides me with access to much of this research. Again, keeping abreast of current developments in my field I feel is an advantage. I do not mind paying for this access. Also, any publicly funded research (e.g. government grants), while not available through the ACM/IEEE digital libraries, is still available online. While some have complained that the ACM/IEEE digital libraries should be open, I don’t entirely agree with this view. That’s a topic for another time.
One of the things that has surprised me as I have gotten further out (in time) from academics is how few professionals maintain their IEEE or ACM memberships. Through these organizations I have had opportunities and created friendships that would not have been possible otherwise. I continue to believe that there is a wonderful opportunity for collaboration between academia and industry, and for me the ACM/IEEE is that conduit. I would strongly encourage anyone out in industry to pick up a membership and take advantage of all the opportunities available to you; conferences, meetups, workshops, etc. It’s some work, but it’s worth it.