A dependency network is a formal logical representation of how system states at one level of a conceptual model are affected by, or dependent on, other antecedent states. Networks represent things such as system states and processes that are relevant to the description of the system being described by the network structure. A single dependency network is hierarchical in structure. However, NetWeaver knowledge bases generally are networks of such hierarchies. In general, a network is composed of other networks whose logical relations to the parent network are defined by relational nodes.
The concept of a truth value comes from the discipline of cognitive science which is basically the science of how we know what we know. In cognitive science the concepts of proposition and truth value are directly related. A proposition is considered to be the smallest unit of thought that can be assigned a measure of truth (ref). NetWeaver knowledge bases are based on propositional logic and every NetWeaver network asserts some proposition concerning the topic it is constructed to evaluate. The key attribute, or state variable, of a network is its truth value which expresses the degree to which antecedent information supports or contradicts the proposition that the network is designed to test.
If all evidence antecedent to a proposition supports the assertion, then the truth value for the network is 1 (completely true). If all evidence antecedent to an assertion is contrary to that assertion, then the truth value for the network is -1 (completely false). If there is no evidence for or against the assertion, then the truth value is 0 (undetermined). Truth values also may be partially true or partially false. Three conditions give rise to this condition in NetWeaver:
As a noun, the term antecedent literally means “something that comes before something else.” An antecedent network, for example, is a network that another network depends upon for its truth value. The term antecedent (the adjective is antecedent) is equally applicable to dependency networks, relational nodes, and data links. An antecedent network, node, or data link logically precedes its dependent network or node in the sense that the value of the antecedent must be known in order to evaluate a network or node that depends on it.
Although the networks of hierarchies in a NetWeaver knowledge base are more general relational structures than simple hierarchies, there are some restrictions on the structure of knowledge bases because they are based on propositional logic. In particular, the NetWeaver inference engine monitors the construction of knowledge bases, and will not allow a dependency structure that implements circular reasoning. For example, it is not permissible for network A to depend on network B if network B depends upon network A because this clearly leads to circular reasoning.
Dependency networks have three basic behaviors: