Papers of the day   All papers

Towards Nonlinear Disentanglement in Natural Datawith Temporal Sparse Coding

Comments

Bethge Lab: Unsupervised learning of disentangled representations is possible! Our approach relies on videos (e.g. below) and assumes that the factors of variation change sparsely across time. Paper: https://arxiv.org/abs/2007.10930 Code: https://github.com/bethgelab/slow_disentanglement [1/6] https://t.co/bQSQVucFko

7 replies, 645 likes


Yash Sharma: My presentation of our work on moving towards nonlinear disentanglement in natural (visual) data as a part of the Talk@Tübingen seminar series @uni_tue is online! Paper: https://arxiv.org/abs/2007.10930 Code: https://github.com/bethgelab/slow_disentanglement Hope it is informative! https://youtu.be/Y4RUlRAx31I

0 replies, 18 likes


Lavanya: Join us for a great deep learning salon with: ◦ @andrewferlitsch – author of Deep Learning Design Patterns, and ◦ @dylanpaiton – author of this paper – https://arxiv.org/abs/2007.10930 When: Tomorrow. Tues, Sep 1, 5 PM PT. Where: http://bit.ly/wb-salon-sep1 #machinelearning #deeplearning

1 replies, 18 likes


Yash Sharma: Check out our work on unsupervised disentanglement via natural transitions!

0 replies, 17 likes


Weights & Biases: Join us for a great deep learning salon with: ◦ @andrewferlitsch – author of Deep Learning Design Patterns, and ◦ @dylanpaiton – author of this paper – https://arxiv.org/abs/2007.10930 When: Tomorrow. Tues, Sep 1, 5 PM PT. Where: http://bit.ly/wb-salon-sep1 #deeplearning #100daysofmlcode

0 replies, 10 likes


Weights & Biases: We are live! https://www.youtube.com/watch?v=_nFN1iowAv8

0 replies, 8 likes


MONTREAL.AI: Towards Nonlinear Disentanglement in Natural Data with Temporal Sparse Coding Klindt et al.: https://arxiv.org/abs/2007.10930 Code: https://github.com/bethgelab/slow_disentanglement #ArtificialIntelligence #DeepLearning #UnsupervisedLearning https://t.co/RJMZ5HSSfu

0 replies, 3 likes


Eli Sennesh: There's a real art to taking a seemingly intuitive inductive bias like "objects retain their properties over time" and turning it into a SOTA neural model. Well done!

0 replies, 3 likes


Charles 🎉 Frye: really cool paper! we'll be bringing author @DylanPaiton on to the @weights_biases salon next month to talk about it

0 replies, 2 likes


arXiv CS-CV: Towards Nonlinear Disentanglement in Natural Data with Temporal Sparse Coding http://arxiv.org/abs/2007.10930

0 replies, 2 likes


Content

Found on Jul 22 2020 at https://arxiv.org/pdf/2007.10930.pdf

PDF content of a computer science paper: Towards Nonlinear Disentanglement in Natural Datawith Temporal Sparse Coding