tel

59 ms·

571.
by tel 10y ago
I often feel like "learning Go" is deceptively a few steps deep. When one learns Chess you effectively can get to the tactical aspect as soon as you know how the pieces move. When you "learn Go" at the first level you un
572.
by tel 11y ago
You can avoid it's use through create-class, but it's freely supported and suggested to use it to access the lifecycle callbacks. Given that there is a supported and suggested pathway to achieving core functionality of the library
573.
by tel 11y ago
That's true and a good point! I'm reading into this as talking to a much broader context. STM in a simple request->response web application probably is overkill and I'm much more aligned with his recommendation now.
574.
by tel 11y ago
It's sad to see that STM is not recommended. It is hugely important in Haskell. I jump to believe that this is an artifact of type-driven STM, but it may be something else instead. I've recently moved from Component to Mount. It
575.
by tel 11y ago
Generics and many kinds of subtyping do cause challenges. This is part of why Scala's type inference is so bad.
576.
by tel 11y ago
Ha, typo :(
577.
Nose.js left-pad is Accidentally Quadratic(accidentallyquadratic.tumblr.com)
3 points by tel 11y ago|0 comments
578.
by tel 11y ago
I hope so too! I'd love to attend!
579.
by tel 11y ago
Without commenting on LambdaConf's statement, PrlConf refers to the JonPrl language but much, much more broadly the general PRL ("Proof Refinement Logic") system of proof languages of which Cornell's NuPRL ( http:/&
580.
by tel 11y ago
Without commenting on whether or not LamdbaConf is actually doing those things, the perception that it is is both (a) out of the control of LambdaConf and (b) completely up to the interpretation of, e.g., the PrlConf organizers. In some s
581.
by tel 11y ago
Typically evaluation of a Go board runs something like (1) Identify all the groups. Most important are any groups larger than about 4-6 stones or groups that cannot be sacrificed for some other reason. You'll often see some straggler g
582.
by tel 11y ago
The biggest most canonical English resource is http://senseis.xmp.net . You can get lost there for days.
583.
by tel 11y ago
Well, existential types are pretty similar. If you've got (exists a . a) then you have to be prepared for it to be a float, but also prepared for it to be a list, or a string, but also a tuple, some type for which we only know that it
584.
by tel 11y ago
I'm not too familiar with SBCL but I highly doubt that those are static types. This sort of information is pretty ascertainable if you really want it, but it's going to be rough going doing non-runtime checking over types this ric
585.
by tel 11y ago
God this is exciting. One of the most painful bits about Scala for me is that documentation cannot be centrally located!
586.
by tel 11y ago
That's the only empirical way, yes. We can also just talk about what it would mean in theory, though. In this case, we'd say that AlphaGo is well trained to the training data set sampling distribution but that may be far from the
587.
by tel 11y ago
All depends on how you define your data sets I suppose. IGS games include some professional games which, if it is the case that AG is perfectly trained on the amateur mode, were smoothed away.
588.
by tel 11y ago
To underscore hyperpape: Tesuji isn't a trick play, it's more like a power play. Each player can read out how a fight is going and see their line far into the future. Two professionals will pick two lines, two suji, which are in
589.
by tel 11y ago
Usually the nomenclature is "W+Resign"
590.
by tel 11y ago
I think it's more that the value network includes moves which look plausible but won't concentrate around the answer to a forcing move as having >99% probability. A human has a heuristic: "I must play here else I lose&quot
591.
by tel 11y ago
No, it plays deep but only so broad. It uses a neural net (which playing by itself without MCTS already beats Pachi with like 80% probability) to sketch out the best moves until the end and then rates each move on its chance of winning. Thi
592.
by tel 11y ago
While AlphaGo has been improving, it's also a little hard to see whether AG playing Fan Hui wasn't already strong enough to defeat LSD. MCTS picks the move most probable to win and if AG gets an early advantage it'll play sla
593.
by tel 11y ago
It plays out thousands of moves all the way until the end of the game using its neural net to quickly sketch optimal play. That's the big advantage! Humans read moves until they feel an outcome is favorable. AlphaGo reads out till endg
594.
by tel 11y ago
I think from a technological perspective there's very little question that AlphaGo could play ko. I would have imagined that AlphaGo would be better at ko than most human players since it's a question of balancing risk across the
595.
by tel 11y ago
I'm very aware of that. What I'm saying is that AlphaGo is not merely a neural net reporting best moves directly off the forward propagation. There are two nets which essentially act as proposal distributions for an exploration&#x
596.
by tel 11y ago
Are they vastly less, though? The core of the algorithm is still a deep Monte Carlo Tree Search which AlphaGo gets quite a boost on computationally for being able to fire it off in parallel. It's obviously incorrect to take the trainin
597.
by tel 11y ago
I'm copying the op who I actually think meant to say "keep track of state", but "stack of state" feels like having a bunch of "pieces of state" all tidily stacked up in the corner so despite the potential
598.
by tel 11y ago
Sorting and filtration are super bread-and-butter FP. The "hard" bits are the IO to get the results and to print them out. Solve the first with a nice HTTP client lib like Wreq, bring in the IO language to do it, finish it off usi
599.
by tel 11y ago
I'll toss it out there that 100% pure FP is pretty much the epitome of keeping a stack of state somewhere... But I wouldn't claim it's too similar or intuitive to someone more used to standard state management.
600.
by tel 11y ago
Not necessarily. It's very possible for one system to embed inside the other. In this case you can often talk about exactly the same axiom/theorem/whatever within different inferential systems. An interesting example is the A

More ›