arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 0903.5342math.PRcs.LGmath.STstat.TH
收藏

无限树上的精确非参数贝叶斯推断

Exact Non-Parametric Bayesian Inference on Infinite Trees

Marcus Hutter

摘要:

注射了。从未知分布的数据,我们考虑预测未来项目的问题。估计概率密度的自适应方法是递归地将域细分为适当的数据相关粒度。贝叶斯将为“细分”分配一个与数据无关的先验概率,这将导致在无限(许多)树上的先验。我们推导出一个精确的,快速的,简单的推理算法,这样的先验,数据证据,预测分布,有效的模型维数,矩,和其他数量。我们证明了渐近收敛和一致性结果,并说明了我们的模型在一些原型函数上的行为。

英文摘要:

Given i.i.d. data from an unknown distribution, we consider the problem of predicting future items. An adaptive way to estimate the probability density is to recursively subdivide the domain to an appropriate data-dependent granularity. A Bayesian would assign a data-independent prior probability to "subdivide", which leads to a prior over infinite(ly many) trees. We derive an exact, fast, and simple inference algorithm for such a prior, for the data evidence, the predictive distribution, the effective model dimension, moments, and other quantities. We prove asymptotic convergence and consistency results, and illustrate the behavior of our model on some prototypical functions.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑