Fork me on GitHub
#off-topic
<
2016-11-22
>
rmoehn02:11:10

@sveri What the network in a reinforcement learner does is estimate the expected reward in a certain state of the game. This is pretty quick: put in a state, out come the actions with their associated expected rewards. If I understand and remember correctly, Alpha Go takes a while for its decision, because it generates possible game trees and evaluates tons of possible states. You can do this if you have time, but you don't have to. You get a trained network by running your training/learning algorithm on many games against itself and against other opponents, but I guess you already knew this part.

rmoehn02:11:55

The roundup lecture of David Silver's RL course might be a good watch: https://www.youtube.com/watch?v=kZ_AUmFcZtk David Silver is one of the guys who built Alpha Go and the Atari games RL algorithm.

sveri09:11:26

@rmoehn Thanks again, that looks like a nice course, it also has pointer to a free book which is even better 🙂

devo17:11:24

Anyone know why comments are disabled on videos posted to the ClojureTV youtube channel?

Alex Miller (Clojure team)17:11:51

they mostly are not disabled, but I think the Euro ones must have had that disabled by default

Alex Miller (Clojure team)17:11:07

I can look at it, but I’ve just been busy with other things

devo17:11:41

@alexmiller no problem, was just curious 🙂

Alex Miller (Clojure team)17:11:06

sometimes the videographers create the playlist and set the default settings in weird ways

Alex Miller (Clojure team)17:11:39

seems to be 100+ comments held in moderation too :)

Alex Miller (Clojure team)17:11:08

@devo ok, should be allowing comments now.

devo17:11:09

sweet, thanks! Still a lot to learn and it's nice having some of the comments there that help expand on the material in the talks.

raynes21:11:34

Ladies and gentleman, I'm happy to announce I'm joining Dollar Shave Club on their data engineering team on December 5th. Super excited. They use every language over the sun in one place or another, including a single Haskell script for something related to docker I'm too excited to recall. I'll be working in Python, PySpark, and I have loads of experience with Scala/Scala Spark and they've often wanted to dig into that as well. The cool thing is that I can likely do some Clojure there as well.

raynes21:11:26

Since I got my offer I've exclusively used their products

raynes21:11:34

OUR products

raynes21:11:04

Sorry for the crazy who gives a crap excitement, I just am so excited to get off contracting gigs and back to a full time job at an awesome place, and to boot I also am reunited with the first frontend developer I ever worked with, Brian Gonzalez at Geni. He's director of frontend there now.

raynes22:11:54

@alexmiller Don't think I won't send you free razors. Their five bladed one is the only blade I've ever been able to use without cutting the hell out of myself because of acne scars. I generally used an electric razor.

raynes22:11:23

My profile pic is pre learning how to shave properly.

raynes22:11:45

Since I'm like this adult thing now who can grow facial hair.