Problem
Consider a template-level CPD where A(U) depends on B(U, V ), allowing for situations where the ground variable A(u) can depend on unbounded number of ground variables B(u, v). As discussed in the text, we can specify the parameterization for the resulting CPD in various ways: we can use a symmetric noisy-or or sigmoid model, or define a dependency of A(u) on some aggregated statistics of the parent set {B(u, v)}. Assume that both A(U) and B(U, V ) are binary-valued. Show that both a symmetric noisy-or model and a symmetric logistic model can be formulated easily using an aggregator CPDs.