The FAIR StarCraft team has a paper at NeurIPS on a convolutional forward model to "defog" a StarCraft map and predict where enemy units will appear.#forwardmodelFTW https://t.co/HP2XYHJe8z
— Yann LeCun (@ylecun) November 30, 2018
The FAIR StarCraft team has a paper at NeurIPS on a convolutional forward model to "defog" a StarCraft map and predict where enemy units will appear.#forwardmodelFTW https://t.co/HP2XYHJe8z
— Yann LeCun (@ylecun) November 30, 2018
Using a temporal ConvNet with dilated convolutions to turn a video of people moving around into a 3D pose sequence.
— Yann LeCun (@ylecun) November 30, 2018
From FAIR-Menlo Park. https://t.co/YzLTUvJB8z
"Robust Artificial Intelligence and Robust Human Organizations" https://t.co/L1oCWhHLph
— Thomas G. Dietterich (@tdietterich) November 29, 2018
Every AI system is deployed by a human organization. In high risk applications, the combined human plus AI system must function as a high-reliability organization to avoid catastrophic errors
Neural networks fooled by unusual poses https://t.co/NE51FX8rNt pic.twitter.com/6o7y4RDCSo
— Carl Vondrick (@cvondrick) November 29, 2018
Their approach can even find groups of neurons that correspond to GAN artifacts, or “mistakes” that GANs make, and remove them. Although I thought adding more artifacts might be more fun than removing them! pic.twitter.com/UeQTgCcEh1
— hardmaru (@hardmaru) November 29, 2018
The specifics are spot-on but I disagree a bit overall. A simple approach solves a problem complex RL algorithms can’t. That’s interesting. Finding pragmatic solutions where current algorithms don’t work leads us to new insights. Not everything has to be model-free end-to-end RL. https://t.co/XATsd6aEnF
— Denny Britz (@dennybritz) November 28, 2018
Insightful critique by @AlexIrpan of the recent Montezuma's Revenge results from Uber AI. https://t.co/MNE6Y6bDH9 He brings up a number of important points regarding the determinism and resetability assumptions the algorithm makes.
— Arthur Juliani (@awjuliani) November 27, 2018
All 41 #NeurIPS2018 papers on #NLProc. You can find the full papers in the proceedings here: https://t.co/PocAoUFCNF https://t.co/LcUSLuAJ0z
— Sebastian Ruder (@seb_ruder) November 27, 2018
Visualizing the Loss Landscape of Neural Nets #NIPS2018
— ML Review (@ml_review) November 27, 2018
By @ljk628
Explores how network architecture affects the loss landscape and its effect of generalization
PyTorchhttps://t.co/qb7F1AorsO
ArXivhttps://t.co/5BTWX9ifNS pic.twitter.com/Dscrlb44rS
Montezuma’s Revenge Solved - Key insights :
— Sebastian Ruder (@seb_ruder) November 26, 2018
1. Remember the exploration frontier.
2. First return to a state, then explore.
3. First solve a problem, then robustify.
4. Random exploration is good enough.
5. Downsample the state or use domain information.https://t.co/P7MXTnIPV5
When A/B tests are not possible, how to estimate the causal effect of online systems? @jakehofman , @duncanjwatts and I present a method that finds natural experiments from log data, whenever the outcome has an independent part. Forthcoming in AOAS! https://t.co/RkbAKziQnm pic.twitter.com/gJIkU7hiGC
— Amit Sharma (@amt_shrma) November 26, 2018
Excited to share the #NeurIPS2018 spotlight video of our paper "A Bayesian Nonparametric View of Count-Min Sketch" with @ryan_p_adams and Michael Mitzenmacher: https://t.co/TwLtJowES3
— Diana Cai (@dianarycai) November 26, 2018