Share This Author
Stein Variational Gradient Descent: A General Purpose Bayesian Inference Algorithm
We propose a general purpose variational inference algorithm that forms a natural counterpart of gradient descent for optimization. Our method iteratively transports a set of particles to match the…
A Kernelized Stein Discrepancy for Goodness-of-fit Tests
A new discrepancy statistic for measuring differences between two probability distributions is derived based on combining Stein's identity with the reproducing kernel Hilbert space theory and a new class of powerful goodness-of-fit tests are derived that are widely applicable for complex and high dimensional distributions.
Breaking the Curse of Horizon: Infinite-Horizon Off-Policy Estimation
A new off-policy estimation method that applies importance sampling directly on the stationary state-visitation distributions to avoid the exploding variance issue faced by existing estimators is proposed.
Variational Inference for Crowdsourcing
By choosing the prior properly, both BP and MF perform surprisingly well on both simulated and real-world datasets, competitive with state-of-the-art algorithms based on more complicated modeling assumptions.
Stein Variational Gradient Descent as Gradient Flow
- Qiang Liu
- Computer ScienceNIPS
- 1 April 2017
This paper develops the first theoretical analysis on SVGD, discussing its weak convergence properties and showing that its asymptotic behavior is captured by a gradient flow of the KL divergence functional under a new metric structure induced by Stein operator.
Communication-efficient Sparse Regression
A communication-efficient approach to distributed sparse regression in the high-dimensional setting and a new parallel and computationally-efficient algorithm to compute the approximate inverse covariance required in the debiasing approach, when the dataset is split across samples.
Stein Variational Policy Gradient
A novel Stein variational policy gradient method (SVPG) which combines existing policy gradient methods and a repulsive functional to generate a set of diverse but well-behaved policies is proposed.
On the Discrimination-Generalization Tradeoff in GANs
This paper shows that a discriminator set is guaranteed to be discriminative whenever its linear span is dense in the set of bounded continuous functions, and develops generalization bounds between the learned distribution and true distribution under different evaluation metrics.
Improving Neural Language Modeling via Adversarial Training
A simple yet highly effective adversarial training mechanism for regularizing neural language models by introducing adversarial noise to the output embedding layer while training the models, allowing for a simple and time efficient algorithm.
Aggregating Ordinal Labels from Crowds by Minimax Conditional Entropy
This work proposes a method to aggregate noisy ordinal labels collected from a crowd of workers or annotators as minimax conditional entropy subject to constraints which encode this observation that workers usually have difficulty distinguishing between two adjacent ordinal classes.