1/ New paper out! @Brain1878 https://t.co/l3VIfg4s9F Robust #MachineLearning of diagnosis from #EEG in disorders of #consciousness powered by @CEAParisSaclay @cdf1530 @Coma_Science @dc_uba @ERC_Research @Inria_Saclay @Inserm @mne_news @NeuroSpin_91 @icm_institute @scikit_learn pic.twitter.com/JsDMUYWw6C
— Denis A. Engemann (@dngman) October 4, 2018
"Causal Embeddings for Recommendation" - really interesting work by @CriteoEng to introduce #causality into embeddings in the context of #recsys #recsys2018 https://t.co/q8mhDZgjUE pic.twitter.com/1ZxLbUmxjW
— Xavier🎗🤖🏃 (@xamat) October 4, 2018
Very interesting take-away on AutoML: " we could have reduced this set
— Andreas Mueller (@amuellerml) October 3, 2018
even further since, in the end, only XGBoost models ended up in the final ensembles for the challenge." https://t.co/dKF481EGBK
Object-Contrastive Networks: Unsupervised Object Representations. ICLR 2019 submission with relevance to the #computervision community. https://t.co/DksoYGZjXE pic.twitter.com/fG4MJS1IOT
— Tomasz Malisiewicz (@quantombone) October 3, 2018
"Interpreting User Inaction in Recommender Systems" - many times we wrongly assume inaction means dislike, this is a great #recsys2018 paper addressing the question of how to model it. pic.twitter.com/vv5yPPRjAu
— Xavier🎗🤖🏃 (@xamat) October 3, 2018
"Explore, Exploit, and Explain: Personalizing Explainable
— Xavier🎗🤖🏃 (@xamat) October 3, 2018
Recommendations with Bandits" - very interesting research by @SpotifyResearch. Bandit approach that has significant improvements as measured in AB tests #recsys2018 pic.twitter.com/idZbIdHHTa
Very interesting first #recsys2018 technical talk on jointly learning recommendations and explanations using a multitask approach combining matrix factorization and adversarial seq2seq. Full text through link here: https://t.co/YGV44MJT0k pic.twitter.com/76XPKAG7ue
— Xavier🎗🤖🏃 (@xamat) October 3, 2018
Really happy with our final #nips2018 ConvAI model!
— Thomas Wolf (@Thom_Wolf) October 3, 2018
We reached a perplexity of 16.3 (very low for LM on open-domain text) & passed 80 in hits@1 (one of the most reliable dialogue metric)
On the privately-held test set!
If you think neural LMs are not good for dialogue—stay tuned! pic.twitter.com/AzHD7rjrn7
Generative Models can be used for so much more than sampling cool images. Check out this paper by Hyunsun Choi and myself on using them for anomaly detection! https://t.co/gn6Wuh3sfV pic.twitter.com/GsDOSPPY2I
— Eric Jang (@ericjang11) October 3, 2018
Nvidia Vid2Vid for high-res video-to-video translation @ctnzr #PyTorch https://t.co/a3G55LjO2K
— Rachel Thomas (@math_rachel) October 2, 2018
SignSGD can provide faster convergence than SGD, similar accuracy, and fault tolerance. @AnimaAnandkumar pic.twitter.com/0ylNJECr0w
— Rachel Thomas (@math_rachel) October 2, 2018
I'm excited to announce our new @PyTorch model predictive control library! We use this in our (forthcoming) NIPS 2018 paper.
— Brandon Amos (@brandondamos) October 2, 2018
This is joint work with I. Jimenez, J. Sacks, B. Boots, and @zicokolter
Project Website: https://t.co/fnx5PQnwP4
GiHub Page: https://t.co/VIw3WmV3vz pic.twitter.com/X5sshxjd97
