Wasserstein Robust Reinforcement Learning
- Mohammed Abdullah, Hang Ren, Jun Wang
- Computer ScienceArXiv
- 30 July 2019
This paper formalises robust reinforcement learning as a novel min-max game with a Wasserstein constraint for a correct and convergent solver following a novel zero-order optimisation method that it believes can be useful to numerical optimisation in general.
Variational f-divergence Minimization
- Mingtian Zhang, Thomas Bird, Raza Habib, Tianlin Xu, D. Barber
- Computer ScienceArXiv
- 27 July 2019
A variational approach that, when combined with the recently introduced Spread Divergence, can be applied to train a large class of latent variable models using any f-divergence.
AFEC: Active Forgetting of Negative Transfer in Continual Learning
- Liyuan Wang, Mingtian Zhang, Yi Zhong
- Computer ScienceNeural Information Processing Systems
- 23 October 2021
This work develops a novel approach named Active Forgetting with synaptic Expansion-Convergence (AFEC), which dynamically expands parameters to learn each new task and then selectively combines them, which is formally consistent with the underlying mechanism of biological active forgetting.
On the Out-of-distribution Generalization of Probabilistic Image Modelling
- Mingtian Zhang, Andi Zhang, Steven G. McDonagh
- Computer ScienceNeural Information Processing Systems
- 4 September 2021
This work proposes a Local Autoregressive model that exclusively models local image features towards improving OOD performance and employs the model to build a new lossless image compressor: NeLLoC (Neural Local Lossless Compressor) and report state-of-the-art compression rates and model size.
Spread Divergence
- Mingtian Zhang, Peter Hayes, Thomas Bird, Raza Habib, D. Barber
- Computer ScienceInternational Conference on Machine Learning
- 21 November 2018
This work demonstrates how to maximize the discriminatory power of a given divergence by parameterizing and learning the spread and gives examples of using a Spread Divergence to train implicit generative models.
Training generative latent models by variational f-divergence minimization
- Mingtian Zhang, Thomas Bird, Raza Habib, Tianlin Xu, D. Barber
- Computer Science
- 27 September 2018
Parallel Neural Local Lossless Compression
- Mingtian Zhang, James Townsend, Ning Kang, David Barber
- Computer ScienceArXiv
- 13 January 2022
This paper proposes two parallelization schemes for local autoregressive models and provides experimental evidence of gains in compression runtime compared to the previous, non-parallel implementation.
Spread Divergences
- D. Barber, Mingtian Zhang, Raza Habib, Thomas Bird
- Computer ScienceArXiv
- 10 December 2019
This work defines a spread divergence on modified p and q and describes sufficient conditions for the existence of such a divergence and demonstrates how to maximize the discriminatory power of a given divergence by parameterizing and learning the spread.
Flow Based Models For Manifold Data
- Mingtian Zhang, Yitong Sun, Steven G. McDonagh, Chen Zhang
- Computer ScienceArXiv
- 29 September 2021
This work proposes to learn a manifold prior that affords benefits to both sample generation and representation quality and an auxiliary benefit of the approach is the ability to identify the intrinsic dimension of the data distribution.
Improving VAE-based Representation Learning
- Mingtian Zhang, Tim Z. Xiao, Brooks Paige, David Barber
- Computer ScienceArXiv
- 28 May 2022
It is shown that by using a decoder that prefers to learn local features, the remaining global features can be well captured by the latent, which significantly improves performance of a downstream classi-cation task.
...
...