Get our free extension to see links to code for papers anywhere online!

Chrome logo Add to Chrome

Firefox logo Add to Firefox
Picture for Michael Arbel

Michael Arbel

UCL

KALE Flow: A Relaxed KL Gradient Flow for Probabilities with Disjoint Support


Jun 16, 2021
Pierre Glaser, Michael Arbel, Arthur Gretton

Figure 1 for KALE Flow: A Relaxed KL Gradient Flow for Probabilities with Disjoint Support
Figure 2 for KALE Flow: A Relaxed KL Gradient Flow for Probabilities with Disjoint Support
Figure 3 for KALE Flow: A Relaxed KL Gradient Flow for Probabilities with Disjoint Support
Figure 4 for KALE Flow: A Relaxed KL Gradient Flow for Probabilities with Disjoint Support

  Access Paper or Ask Questions

Annealed Flow Transport Monte Carlo


Feb 15, 2021
Michael Arbel, Alexander G. D. G. Matthews, Arnaud Doucet

Figure 1 for Annealed Flow Transport Monte Carlo
Figure 2 for Annealed Flow Transport Monte Carlo

  Access Paper or Ask Questions

Deep Reinforcement Learning with Dynamic Optimism


Feb 09, 2021
Ted Moskovitz, Jack Parker-Holder, Aldo Pacchiano, Michael Arbel

Figure 1 for Deep Reinforcement Learning with Dynamic Optimism
Figure 2 for Deep Reinforcement Learning with Dynamic Optimism
Figure 3 for Deep Reinforcement Learning with Dynamic Optimism
Figure 4 for Deep Reinforcement Learning with Dynamic Optimism

  Access Paper or Ask Questions

The Unreasonable Effectiveness of Patches in Deep Convolutional Kernels Methods


Jan 19, 2021
Louis Thiry, Michael Arbel, Eugene Belilovsky, Edouard Oyallon

Figure 1 for The Unreasonable Effectiveness of Patches in Deep Convolutional Kernels Methods
Figure 2 for The Unreasonable Effectiveness of Patches in Deep Convolutional Kernels Methods
Figure 3 for The Unreasonable Effectiveness of Patches in Deep Convolutional Kernels Methods
Figure 4 for The Unreasonable Effectiveness of Patches in Deep Convolutional Kernels Methods
* International Conference on Learning Representation (ICLR 2021), 2021, Vienna (online), Austria 
  Access Paper or Ask Questions

Efficient Wasserstein Natural Gradients for Reinforcement Learning


Oct 12, 2020
Ted Moskovitz, Michael Arbel, Ferenc Huszar, Arthur Gretton

Figure 1 for Efficient Wasserstein Natural Gradients for Reinforcement Learning
Figure 2 for Efficient Wasserstein Natural Gradients for Reinforcement Learning
Figure 3 for Efficient Wasserstein Natural Gradients for Reinforcement Learning
Figure 4 for Efficient Wasserstein Natural Gradients for Reinforcement Learning

  Access Paper or Ask Questions

Estimating Barycenters of Measures in High Dimensions


Jul 14, 2020
Samuel Cohen, Michael Arbel, Marc Peter Deisenroth

Figure 1 for Estimating Barycenters of Measures in High Dimensions
Figure 2 for Estimating Barycenters of Measures in High Dimensions
Figure 3 for Estimating Barycenters of Measures in High Dimensions
Figure 4 for Estimating Barycenters of Measures in High Dimensions
* In submission 
  Access Paper or Ask Questions

A Non-Asymptotic Analysis for Stein Variational Gradient Descent


Jun 17, 2020
Anna Korba, Adil Salim, Michael Arbel, Giulia Luise, Arthur Gretton

Figure 1 for A Non-Asymptotic Analysis for Stein Variational Gradient Descent

  Access Paper or Ask Questions

Synchronizing Probability Measures on Rotations via Optimal Transport


Apr 01, 2020
Tolga Birdal, Michael Arbel, Umut ĹžimĹźekli, Leonidas Guibas

Figure 1 for Synchronizing Probability Measures on Rotations via Optimal Transport
Figure 2 for Synchronizing Probability Measures on Rotations via Optimal Transport
Figure 3 for Synchronizing Probability Measures on Rotations via Optimal Transport
Figure 4 for Synchronizing Probability Measures on Rotations via Optimal Transport
* Accepted for publication at CVPR 2020, includes supplementary material. Project website: https://github.com/SynchInVision/probsync 
  Access Paper or Ask Questions

KALE: When Energy-Based Learning Meets Adversarial Training


Mar 10, 2020
Michael Arbel, Liang Zhou, Arthur Gretton

Figure 1 for KALE: When Energy-Based Learning Meets Adversarial Training
Figure 2 for KALE: When Energy-Based Learning Meets Adversarial Training
Figure 3 for KALE: When Energy-Based Learning Meets Adversarial Training

  Access Paper or Ask Questions

Kernelized Wasserstein Natural Gradient


Oct 25, 2019
Michael Arbel, Arthur Gretton, Wuchen Li, Guido Montufar

Figure 1 for Kernelized Wasserstein Natural Gradient
Figure 2 for Kernelized Wasserstein Natural Gradient
Figure 3 for Kernelized Wasserstein Natural Gradient
Figure 4 for Kernelized Wasserstein Natural Gradient

  Access Paper or Ask Questions

Maximum Mean Discrepancy Gradient Flow


Jun 11, 2019
Michael Arbel, Anna Korba, Adil Salim, Arthur Gretton

Figure 1 for Maximum Mean Discrepancy Gradient Flow
Figure 2 for Maximum Mean Discrepancy Gradient Flow

  Access Paper or Ask Questions

On gradient regularizers for MMD GANs


Oct 27, 2018
Michael Arbel, Dougal J. Sutherland, Mikołaj Bińkowski, Arthur Gretton

Figure 1 for On gradient regularizers for MMD GANs
Figure 2 for On gradient regularizers for MMD GANs
Figure 3 for On gradient regularizers for MMD GANs
Figure 4 for On gradient regularizers for MMD GANs
* Code available at https://github.com/MichaelArbel/Scaled-MMD-GAN . v2: NIPS camera-ready version 
  Access Paper or Ask Questions

Kernel Conditional Exponential Family


Apr 08, 2018
Michael Arbel, Arthur Gretton

Figure 1 for Kernel Conditional Exponential Family
Figure 2 for Kernel Conditional Exponential Family
Figure 3 for Kernel Conditional Exponential Family
Figure 4 for Kernel Conditional Exponential Family

  Access Paper or Ask Questions

Demystifying MMD GANs


Mar 21, 2018
Mikołaj Bińkowski, Dougal J. Sutherland, Michael Arbel, Arthur Gretton

Figure 1 for Demystifying MMD GANs
Figure 2 for Demystifying MMD GANs
Figure 3 for Demystifying MMD GANs
Figure 4 for Demystifying MMD GANs
* Published at ICLR 2018: https://openreview.net/forum?id=r1lUOzWCW . v4: actually-final version: non-existence of unbiased estimators for IPMs and FID; clarity edits to the main proof 
  Access Paper or Ask Questions

Efficient and principled score estimation with Nyström kernel exponential families


Mar 13, 2018
Dougal J. Sutherland, Heiko Strathmann, Michael Arbel, Arthur Gretton

Figure 1 for Efficient and principled score estimation with Nyström kernel exponential families
Figure 2 for Efficient and principled score estimation with Nyström kernel exponential families
Figure 3 for Efficient and principled score estimation with Nyström kernel exponential families
* v5: Final version to be published at AISTATS 2018 
  Access Paper or Ask Questions