Variational Inference
PulseAugur coverage of Variational Inference — every cluster mentioning Variational Inference across labs, papers, and developer communities, ranked by signal.
-
New VPR method improves Bayesian posterior sampling accuracy
Researchers have introduced Variational Predictive Resampling (VPR), a new method designed to improve the accuracy of Bayesian posterior sampling. VPR leverages variational inference's predictive capabilities within a r…
-
New neural tilting framework improves AI safety inference
Researchers have developed a new neural exponential tilting framework for variational inference in Lévy-driven stochastic differential equations. This method addresses the intractability of Bayesian inference for proces…
-
New AI research explores advanced methods for uncertainty estimation and Bayesian inference
Researchers have developed a new variational Bayesian framework that directly targets the posterior-predictive distribution, jointly learning approximations for both the posterior and predictive distributions. This appr…
-
新论文从单一KL恒等式推导出指数族结果
研究人员发现了一个指数族的基本恒等式，指数族是现代机器学习技术（如softmax和高斯分布）的关键分布。该恒等式简化了变分推断和强化学习中几个关键结果的推导，包括勾股定理和吉布斯变分原理。这些研究结果在一个独立的笔记中提出，为理解这些复杂的数学概念提供了一种更简化的方法。
-
Symmetry Guarantees Statistic Recovery in Variational Inference
Two new papers explore how symmetries in target distributions can guarantee the recovery of certain statistics during variational inference, even when the chosen variational family is misspecified. The research provides…