Will Grathwohl
Will Grathwohl
Research Scientist, Deepmind
Verified email at - Homepage
Cited by
Cited by
Ffjord: Free-form continuous dynamics for scalable reversible generative models
W Grathwohl, RTQ Chen, J Bettencourt, I Sutskever, D Duvenaud
arXiv preprint arXiv:1810.01367, 2018
Invertible residual networks
J Behrmann, W Grathwohl, RTQ Chen, D Duvenaud, JH Jacobsen
International Conference on Machine Learning, 573-582, 2019
Your classifier is secretly an energy based model and you should treat it like one
W Grathwohl, KC Wang, JH Jacobsen, D Duvenaud, M Norouzi, ...
arXiv preprint arXiv:1912.03263, 2019
Backpropagation through the void: Optimizing control variates for black-box gradient estimation
W Grathwohl, D Choi, Y Wu, G Roeder, D Duvenaud
arXiv preprint arXiv:1711.00123, 2017
Learning the Stein Discrepancy for Training and Evaluating Energy-Based Models without Sampling
W Grathwohl, KC Wang, JH Jacobsen, D Duvenaud, R Zemel
International Conference on Machine Learning, 2020
Deep reinforcement learning and simulation as a path toward precision medicine
BK Petersen, J Yang, WS Grathwohl, C Cockrell, C Santiago, G An, ...
Journal of Computational Biology 26 (6), 597-604, 2019
Oops i took a gradient: Scalable sampling for discrete distributions
W Grathwohl, K Swersky, M Hashemi, D Duvenaud, C Maddison
International Conference on Machine Learning, 3831-3841, 2021
Understanding the limitations of conditional generative models
E Fetaya, JH Jacobsen, W Grathwohl, R Zemel
arXiv preprint arXiv:1906.01171, 2019
Disentangling space and time in video with hierarchical variational auto-encoders
W Grathwohl, A Wilson
arXiv preprint arXiv:1612.04440, 2016
Joint energy-based models for semi-supervised classification
S Zhao, JH Jacobsen, W Grathwohl
ICML 2020 Workshop on Uncertainty and Robustness in Deep Learning 1, 2020
Gradient-based optimization of neural network architecture
W Grathwohl, E Creager, SKS Ghasemipour, R Zemel
Optimal design of stochastic DNA synthesis protocols based on generative sequence models
EN Weinstein, AN Amin, WS Grathwohl, D Kassler, J Disset, D Marks
International Conference on Artificial Intelligence and Statistics, 7450-7482, 2022
No MCMC for me: Amortized samplers for fast and stable training of energy-based models
D Duvenaud, J Kelly, K Swersky, M Hashemi, M Norouzi, W Grathwohl
Score-based diffusion meets annealed importance sampling
A Doucet, W Grathwohl, AGDG Matthews, H Strathmann
arXiv preprint arXiv:2208.07698, 2022
Directly training joint energy-based models for conditional synthesis and calibrated prediction of multi-attribute data
J Kelly, R Zemel, W Grathwohl
arXiv preprint arXiv:2108.04227, 2021
No conditional models for me: Training joint ebms on mixed continuous and discrete data
J Kelly, WS Grathwohl
Energy Based Models Workshop-ICLR 2021, 2021
Graph generation with energy-based models
J Liu, W Grathwohl, J Ba, K Swersky
ICML Workshop on Graph Representation Learning and Beyond (GRL+), 2020
Continuous diffusion for categorical data
S Dieleman, L Sartran, A Roshannai, N Savinov, Y Ganin, PH Richemond, ...
arXiv preprint arXiv:2211.15089, 2022
Self-conditioned embedding diffusion for text generation
R Strudel, C Tallec, F Altché, Y Du, Y Ganin, A Mensch, W Grathwohl, ...
arXiv preprint arXiv:2211.04236, 2022
Training Glow with constant memory cost
X Li, W Grathwohl
NIPS Workshop on Bayesian Deep Learning, 2018
The system can't perform the operation now. Try again later.
Articles 1–20