34.
hardmaru
@hardmaru
Our work on "World Models" is outοΌ
Can neural network agents dream, and learn inside of their own dream worlds?β¦ twitter.com/i/web/status/9β¦
Can neural network agents dream, and learn inside of their own dream worlds?β¦ twitter.com/i/web/status/9β¦
35.
hardmaru
@hardmaru
#ObjectDetection with Faster R-CNN. Here's a detailed explanation of how this state-of-the-art model works.
tryolabs.com/blog/2018/01/1β¦
tryolabs.com/blog/2018/01/1β¦
36.
37.
38.
39.40.
hardmaru
@hardmaru
until a few days ago it made more sense to use extra GPUs to mine bitcoins rather than for tuning hyperparameters for some model
hardmaru
@hardmaru
"Born Again Neural Networks"
By distilling a teacher model to a student model with an identical architecture, thβ¦ twitter.com/i/web/status/9β¦
By distilling a teacher model to a student model with an identical architecture, thβ¦ twitter.com/i/web/status/9β¦
hardmaru
@hardmaru
#NIPS2017 ali ramani's the test of time talk was amazing!
Paraphrasing: "We apply brittle optimization techinqβ¦ twitter.com/i/web/status/9β¦
Paraphrasing: "We apply brittle optimization techinqβ¦ twitter.com/i/web/status/9β¦
41.
hardmaru
@hardmaru
Deep RL has a lot of catching up to do. youtu.be/fRj34o4hN4I pic.twitter.com/mOxwyJBt9S
...but wait! There's more!
34.
fakhright
@fakhright
astaghfirullah peng.krim guaaaaaaaa..............a *salto sambil solat*
