Oct 2022· International Conference on Learning Representations· Vol abs/2210.00580· 120 citations· ⚡ 9 influential· 65 references
Computer ScienceMathematics
Abstract
This paper builds bridges between two families of probabilistic algorithms: (hierarchical) variational inference (VI), which is typically used to model distributions over continuous spaces, and generative flow networks (GFlowNets), which have been used for distributions over discrete structures such as graphs. We demonstrate that, in certain cases, VI algorithms are equivalent to special cases of GFlowNets in the sense of equality of expected gradients of their learning objectives. We then point out the differences between the two families and show how these differences emerge experimentally. Notably, GFlowNets, which borrow ideas from reinforcement learning, are more amenable than VI to off-policy training without the cost of high gradient variance induced by importance sampling. We argue that this property of GFlowNets can provide advantages for capturing diversity in multimodal target distributions.
Combining predictions from different models can improve performance at machine learning tasks, but the training of the individual models and the rule used to combine them are typically chosen separately, and by ad hoc means. Recent advances in distributional optimisation (i.e. where the optimisation occurs over the set...
Cong-Ye Wang, Yan-Kai Lin, Zhe-Yang Shen et al.· 0 citations
This work describes an automatic criterion for full-state rejuvenation of the Gibbs sampler, derived from the Gelman-Rubin statistic, which plays a key role in speeding up learning convergence.
Tiago da Silva, Diego Mesquita, S. Lahlou· 0 citations
DENSE is an ensemble of multi layer perceptrons (MLPs) that is trained to minimize a modified Cram´er–von Mises distance in the joint space of inputs and outputs that enables non-parametric representations of arbitrary densities without (complex) likelihood design.
Leon Winheim, Hayk Amirkhanian, MarkusWalker et al.· 0 citations
In this thesis I develop methods for statistical inference when the distributions arising from complex biological systems are multi-modal, geometrically structured, and sometimes only defined up to a normalizing constant. I start from variational inference and, when analytic update equations are unavailable, move to bl...
It is demonstrated that the Markovian assumption can both hamper signal propagation during training and catastrophically reduce the learned sampler's expressivity due to state aliasing.
Tiago da Silva, Esmeralda S. Whitammer, S. Lahlou· 0 citations
This work provides an elementary derivation of a single-sample post-stratified estimator—previously derived via transport equations—and proves a variance reduction relative to simple random sampling, expanding the applicability of implicit reparameterization and reducing its computational complexity.
Learn how Google DeepMind and GovTech Singapore uses lightweight on-device Gemma models to intercept conversational scams in private chats with zero cloud data egress.
