eutectic

27 ms·

241.
by eutectic 9y ago
There is a value network which estimates the win rate for the current player from a given board state, and a policy network which estimates the probability that each move should be played. As of the more recent iterations, these networks sh
242.
by eutectic 9y ago
Reverse sear is even better; by going oven-first you dry out the surface and get a better sear.
243.
by eutectic 9y ago
Why only consider within-country inequality (and, yes, I realize that the answer is 'because it's partly a publicity stunt, and humans are biased and parochial')? The average American is incomparable wealthy compared to the a
244.
by eutectic 9y ago
Probably this is a consequence of least-squares training; the 'average' solution is extremely unlikely.
245.
by eutectic 9y ago
This is as close to magic as humanly possible.
246.
by eutectic 9y ago
You are living in a bubble if you think that the balance of power between employers and employees is mostly determined by how easy it is to start a business.
247.
by eutectic 9y ago
Since when was terseness a bad thing?
248.
by eutectic 9y ago
I remember seeing a demonstration by a deep learning guy of a commercially available robot cleaning a house under remote control. You are seriously underestimating the difficulty of developing software to solve these problems in an integrat
249.
by eutectic 9y ago
There is a small fraction puzzles for which simple backtracking gets stuck on unproductive branches and essentially never finishes.
250.
by eutectic 9y ago
It is important to realise that there is a big difference between the value to the individual and the value to society of a university degree. Companies use tertiary education as a proxy for IQ and conscientiousness (or willingness to toler
251.
by eutectic 9y ago
Alphago used several hand-crafted features as of the Nature paper, so DeepMind at least is not above a little feature engineering.
252.
by eutectic 9y ago
I suspect you might be able to do surprisingly well with just a few simple features, e.g. what did I last see at each position and how long ago was that, how many of each enemy unit have I seen simultaneously and at what time, etc. As to th
253.
by eutectic 9y ago
You could have made exactly that this argument about flight before the Wright brothers.
254.
by eutectic 9y ago
And 1600x worse by kinetic energy.
255.
by eutectic 9y ago
With 7.5 billion people and counting, we have to be realistic about what it takes to feed everyone.
256.
by eutectic 9y ago
Once you are used to them roundabouts are for the most part pleasant to use, and safer than intersections. It's especially nice not to have to turn across incoming traffic.
257.
by eutectic 9y ago
I was talking more about cryptographic hardness assumptions such as the existence of one-way functions. Nevertheless, even in http://www.ccs.neu.edu/home/wichs/thesis.pdf , which was the first search result for tha
258.
by eutectic 9y ago
Even crypto algorithms rely on unproven assumptions for their security, even ignoring e.g. side-channel attacks.
259.
by eutectic 9y ago
Can you please expand a bit on how it worked?
260.
by eutectic 9y ago
A variable becomes '1' in the corresponding position and '.' everywhere else, and similarly with negations of variables and '0'. The regex is then just an alternation of these sub-regexes. This incurs just a li
261.
by eutectic 9y ago
This problem is NP-hard by reduction from SAT. Treat each column as a truth variable and use the rows to encode CNF clauses. For example, `(A | ^C)` becomes `(1..)|(..0)`. Then set all the column regexes to `( 0* )|( 1* )` to enforce a cons
262.
by eutectic 9y ago
You could expand each regex to a regex on the whole table, and then take the intersection of the corresponding NFA/DFAs. Unfortunately, I suspect this takes exponential (or worse?) time in the worst case.
263.
by eutectic 9y ago
It's both.
264.
by eutectic 9y ago
When you perform quicksort/radix sort you can identify the range of each partition as you go. You could then use this to inform your search, e.g. there's no point in searching within an interval of width <K.
265.
by eutectic 9y ago
Thanks for the correction, my bad. The technical term for the kind of thing I was thinking of is 'multiscale composites', and there already seems to be a lot written about carbon fiber/carbon nanotube mixed composites which w
266.
by eutectic 9y ago
This is a fantastic example of how to do math/science journalism right. It's not afraid to get into the technical details, but everything is explained so that someone without much math background can understand. The human interest
267.
by eutectic 9y ago
Carbon fiber works so well in composites because it is possible to create long, continuous fibers which can be woven into cloth for directional reinforcement. Maybe it is possible to use the strengths of each by adding CNCs to carbon compos
268.
by eutectic 9y ago
High-level is not a synonym for interpreted, dynamic.
269.
by eutectic 9y ago
My understanding is that most of the development effort has been moved over.
270.
by eutectic 9y ago
Daala is dead, AV1 is the current contender.

More ›