### B. the same conditional probability distribution. We

B. THE

BAYESIAN LEARNING MODULE:

Methods for learning Bayesian networks

can discover dependency structure between observed variables. Although these

methods are useful in many applications, they run into computational and

statistical problems in domains that involve a large number of variables. we

consider a solution that is applicable when many variables have similar

behavior. We introduce a new class of models, module networks, that explicitly

partition the variables into modules that share the same parents in the network

and the same conditional probability distribution. We define the semantics of

module networks, and describe an algorithm that learns the modules’ composition and their

dependency structure from data. Evaluation on real data in the domains of gene

expression and the stock market shows that module networks generalize better

than Bayesian networks, and that the learned module network structure reveals

regularities that are obscured in learned Bayesian networks.

4.2.1 Generating

Initial Causal Pathways

This subsection first introduces the

representation of causal pathways in the ST space, and then elaborates how to

generate initial causal pathways.

Gaussian Bayesian Network (GBN). GBN is a special form of Bayesian

network for probabilistic inference with continuous Gaussian variables in a

DAG, in which each variable is assumed as linear function of its parents 9.

The ST causal relations of air pollutants are encoded in a GBN-based graphical

model, to represent both local and ST dependencies. Here we choose GBN to model

the causalities because: 1) GBN provides a simple way to represent the

dependencies among multiple pollutants variables, both locally and in the ST

space. 2) GBN models continuous variables rather than discrete values. Due to

the sensors monitor the concentration of pollutants per hour; GBN could help

better capture the fine-grained knowledge through the dependencies of these

continuous values.

In

this subsection, in light of the extricated coordinated examples also,

applicant sensors from the example digging module for each poison Pcmsn, we utilize Pcmsn

to speak to consistent esteems in the graphical model. 3) The qualities of

urban information fit the GBN show well. As appeared in Fig. 7, the

dissemination of 1-hour distinction (current esteem short the esteem 1-hour

back) of air toxins and meteorological information obey Gaussian appropriation.

In the following areas, standardized 1-hour contrasts of time arrangement

information will be utilized as contributions for the model.