Arc States
Context
- An arc connects a Parent Node with a Child Node, and it represents a direct probabilistic relationship between those two linked nodes.
- Creating a new arc between nodes (see Arc Creation Mode) generates a new entry in the probability table associated with the Child Node.
Arc States
An arc can be in one of four states:
-
Normalis the default state of any arc. - It appears as a solid line with an arrowhead.
-
Selected: By left-clicking on an unselected arc, it becomes Selected. Then, you can perform a number of operations on the arc, such as Delete, Change Orientation (see Arc Contextual Menu). - A Selected arc is highlighted with a thicker line and a bigger arrowhead.
-
Fixed Arcmeans that prior knowledge is applied to the network structure before performing learning. - The line of a Fixed Arc is dotted.
-
A Temporal Arc indicates a temporal relationship between a parent node N at time step t, denoted N(t), and a strictly identical child node at time step t+1, denoted N(t+1) - A Temporal Arc transforms the network into a dynamic Bayesian network. - The parent node in a temporal relationship can only have one temporal child node. - A Temporal Arc is shown with a dashed line.
Arc Appearance
Translucence
In addition to the formal Arc States described above, an arc can appear translucent:
There are a number of possible reasons for this particular look of the arc:
- One or both of the nodes that are connected by the arc are excluded from display because of the Class Filter.
- The arc does not meet the threshold defined in one of the Visual Arc Analysis functions:
- Arc Force or Kullback-Leibler Divergence
- Arc Mutual Information
- Arc Pearson's Correlation
- Regardless of the reason, while an arc is translucent, it cannot be selected or otherwise modified.
Highlight
Specific functions can highlight an arc with distinct colors, such a blue and pink:
Among these functions are:
Undirected Arcs
Arcs can appear as undirected in special contexts:
However, the functions that produce this undirected appearance are very different in nature.
- The Display Skeleton function simply removes all arrowheads in the network from display, for instance when explaining a graph to someone who is not familiar with Bayesian networks. In non-causal Bayesian networks, a casual observer may misinterpret a directed arc as a causal relationship. Temporarily removing all arrowheads reduces the risk of a misunderstanding.
- The Equivalence Classes function visualizes the Equivalence Class by displaying the Essential Graph for the given network. If the Essential Graph contains Edges, i.e., undirected arcs, they are shown as such.
Note the important difference between Display Skeleton and Equivalence Classes. The former is merely a visual effect while the latter is an analysis function that can provide important structural insights into the network graph.
Icons on Arcs
The information icon attached to the middle of an arc indicates that an Arc Comment is associated with the arc.
If Show Information While Hovering is active, as indicated in the Toolbar by the icon, you will see the Arc Comment when your cursor hovers over the information icon . You can also activate this mode with the I shortcut.