If you are up early tomorrow, come to my talk at #CVPR2019 workshop on Vision With Biased or Scarce Data https://wvbsd.github.io/2019/ I will show our recent results on detecting hard examples using a new angular distance measure. Also talk about many ways to reduce our data reqs
-
Show this thread
-
Co-authors
@BeidiChen@animesh_garg@Anshumali_ Zhiding Weiyung on detecting hard examples@NvidiaAI Active learning with@HuPeiyun@zacharylipton Deva Ramanan. Crowdsourcing with Ashish Khetan@zacharylipton Semi-supervised learning with neural rendering@TanNguyen689 Nhat Ho..1 reply 0 retweets 6 likesShow this thread -
Otherwise you can attend my talk at 10:30AM tomorrow at
#CVPR2019 workshop on semantic information https://sites.google.com/view/wsi-2019/ on Beyond Black-Boxes: Embedding Semantic Structures into Learning1 reply 2 retweets 8 likesShow this thread -
#tensor operations in#NeuralNetworks lead to compression@JeanKossaifi@zacharylipton@furlanel@arankhanna#tensor-train LSTMs are much better for long-term#forecasting@yuqirose@StephanZheng@yisongyue + video prediction Jiahao Wonmin@furongh@jankautz@NvidiaAI1 reply 1 retweet 5 likesShow this thread -
Other ways to add semantic structure into learning: blending symbolic expressions with data and learning common embeddings
@ForoughArabsha1 Sameer Singh. Adding learning to control while retaining stability:@yisongyue@GuanyaShi Anqi Liu Soon Jo Chung@kazizzad@yuqirose1 reply 0 retweets 7 likesShow this thread
Lastly, important to study domain adaptation. We study label shift where label proportions change. More relevant in modern #AI applications (e.g. in cloud services, disease prediction). New generalization bounds for this problem. @kazizzad Anqi Liu Fanny Yang
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.