ted_dunning

38 ms·

601.
by ted_dunning 10y ago
My calendar agrees with you ... we apparently had lunch in December, 2007. Sadly, I don't remember it off-hand. On the other hand, one reason is that I invited a lot of students to come visit (and a number to intern with us).
602.
by ted_dunning 10y ago
Thompson sampling does not need to assume stability. You can inject time features into the model if you want to model seasonality (or, more accurately, ignorances of seasonality) and you can also have a hidden random-walk variable. Yes, if
603.
by ted_dunning 10y ago
This is yet another article that ignores the fact that there is a MUCH better approach to this problem. Thompson sampling avoids the problems of multiple testing, power, early stopping and so on by starting with a proper Bayesian approach.
604.
by ted_dunning 10y ago
That critique had a lot of observations of the form "lisp is less good because I am not used to it". From the point of view of something with more lisp experience, quite a number of these observations would go the other way.
605.
by ted_dunning 10y ago
What is it that worried you? That the team is going to check things over to make sure that they have things done right? The only alternative to that is to just kind of pretend that it all works without doing the hard work to verify. You saw
606.
by ted_dunning 10y ago
Dyn's status page still says they are working the problem. Why would Verizon block access to a downed service?
607.
by ted_dunning 10y ago
MapR | San Jose, CA | Full-Time | Onsite | US Citizen, Green Card other work visa We build the most advanced big data platform around. If you want to work on systems software at a level found in very few places, you should check us out. Ema
608.
by ted_dunning 11y ago
Hmmm... you don't seem quite up to date on Silicon Valley. You seem to forget the older achievements as well. Intel (you seem impressed with them) are right in the heart. Cisco, same. Google seems to be unimpressive to you, but they ar
609.
by ted_dunning 11y ago
Count-min sketches can be adapted to use exponential decay. Presumably any sort of recurrent decay could be used to get approximately rectangular windows as well.
610.
by ted_dunning 11y ago
Intel has had good results with formal methods. This is pretty important to all of the rest of us. Here are some slightly dated references to slides on the topic: http://www.cl.cam.ac.uk/~jrh13/slides/nasa-14apr10&
611.
by ted_dunning 11y ago
There is an inspiration lineage from Dremel, but there are also very substantial differences. Dremel handles nested data, but doesn't like variable schema and tends to generate code early (AFAIK). Drill handles data that changes shape
612.
by ted_dunning 11y ago
There are a number of companies with employees who are involved in Drill. These include MapR, Hortonworks and Twitter. That doesn't imply corporate sponsorship of Drill since all contributors in Apache contribute as individuals.
613.
by ted_dunning 11y ago
Apache has a number of vice-presidents. The reason is that making a release establishes that the Apache Software Foundation is standing behind the release and the licensing. The policy is that this requires the imprimatur of a corporate of
614.
by ted_dunning 11y ago
@polskibus, Heterogeneous can mean different things. Drill attacks problems where there is no well defined schema. Other heterogeneous data source tools are more about accessing different data sources, all relational. To do what Drill does
615.
by ted_dunning 12y ago
Yeah... shameful. I am sure that there is something simple that all the drive manufacturers are just overlooking in their process that would take them to 10^-17 BER. When was the last time you created a physical artifact for sale? That