用有向迴圈圖形模型模擬離散介入資料

我們概括了離散多變數分佈的全域性歸一化介入勢函式表示。

We outline a representation for discretemultivariate distributions in terms of interventional potential functions thatare globally normalized.

該表示式可以用於介入效果的建模，並且該模型中編碼的獨立屬性可以表示為允許迴圈的有向圖。

This representation can be used to modelthe effects of interventions, and the independence properties encoded in thismodel can be represented as a directed graph that allows cycles.

除了用該表示式討論推理和取樣之外，我們給出了一種指數族引數化形式，它允許引數估計被描述為一個凸優化問題；我們還給出了一種採用群L1正則化的引數和結構同時學習的任務凸鬆弛模型。

In addition to discussing inference andsampling with this representation, we give an exponential familyparametrization that allows parameter estimation to be stated as a convexoptimization problem; we also give a convex relaxation of the task of simultaneousparameter and structure learning using group L1-regularization.

該模型用於評估模擬資料和細胞內流式細胞計資料。

The model is evaluated on simulated dataand intracellular flow cytometry data.

圖形模型為表示多元分佈的獨立性提供了一個方便的研究框架。

Graphical models provide a convenientframework for representing independence properties of multivariatedistributions.

最近人們非常關注使用圖形模型對帶有介入措施的資料進行建模，即其中一些變數是通過實驗設定的資料。

There has been substantial recent interestin using graphical models to model data with interventions, that is, data wheresome of the variables are set experimentally.

……

一個與該文相關的網站供參考：

下載英文原文地址：

更多精彩文章請關注微訊號：