Show 

Glossary  S - Z

 

Select link:  In order to do some operations on a link, you first select it by clicking once on it, and it will then be drawn using negative colors to hilite it.  To select more than one link, or remove links from your selection, hold down the ctrl key while clicking on them.  more info

Select node:  In order to do some operations on a node, you first select it by clicking once on it, and it will then be drawn using negative colors to hilite it.  You can select several nodes at a time by clicking on the background and dragging the selection rectangle to include them.  To add or remove nodes from your selection, hold down the ctrl key while you do the selection operation, or right-click on the node and choose Select/Deselect.  more info

Singly-connected:  A singly-connected network is a directed or undirected network that has at most one undirected path between any two nodes.  In other words, it has no loops.  more info

Standard deviation:  The standard deviation is the square root of the variance.  Its units are the same as those of X, and it is a measure of how “spread out” or “imprecise” the distribution is.  more info

State index:  Any state of a node may be referred to by its state index, which is an integer that starts at 0 for the first state, and is one higher for each successive state.  Note that state index is different from state number, and so to distinguish it a # symbol is often placed in front.  For example, if a node has 3 states, the state indexes would be:  #0   #1   #2.

State number:  Any discrete node may have associated with each of its states a number (integer or real) called its state number, which may be used to identify that state in databases or case files, or may be used to provide an "output value" for that state in equations.  Continuous nodes don't have state numbers, but if they have been discretized, they have state thresholds instead.  Note that state number is different from state index .  more info

State thresholds:  Any continuous variable (i.e. continuous node) may be discretized, a process in which a set of numeric borders is used to partition the range of the variable into intervals called states, so that each state corresponds to a range of values.  The numeric values that divide adjacent states are known as thresholds, so the range of a state is from its lower threshold to its upper threshold, with the upper threshold being the same as the lower threshold of the next state.  If reference is made to a state's threshold without specifying lower or upper, then lower is implied.  Only continuous nodes can be discretized, but numbers may also be attached to the states of any discrete node, in which case each state has a single number known as its state number.  See also state index.

State values: See state number above.

States:  A discrete variable can take on one of several values, and these values are called states.  For example the states may be “female, male”, or they might be “US, Europe, Japan, China”, or “True, False”.  With Netica you can just let the states of a variable be numbered, but usually you give them meaningful names.

Structure learning:  Structure learning is the automatic discovery of the link structure of a Bayes net from case data.  Compare with parameter learning.

Support:  The support of a probability distribution is the range of its variable(s) over which the pdf is non zero.  In other words, those values of the variable which are possible. In a Netica equation, it is legal to pass values to a built-in distribution function outside of that distribution's support; the function will then just return 0 for the probability.

Tab delimited text file:  Tab delimited text file is a commonly used term for a form of case file in which the names of the variables appear on the first line, and then below are all the cases (i.e. records), with each case on a single line and having a value for each of the variables, and with all the values and variables in text form and separated by tab characters.  See also CSV file.

Table dialog box:  To change or view the relationship of a node with its parents that is, edit its conditional probability table (CPT) or function table, you use a table dialog box. To obtain select the node, and then choose Table View/Edit or click the image\RelnTOOL.gif toolbar button.  It is not to change the node's basic properties; for that, use a node dialog box instead.  more info

Table selector:  Within the table dialog box, you will find a table selector, which looks like this, and which provides a menu for you to select the type of table that you view.    

Target Node: A target node is a node whose beliefs we want to know after belief updating.  Also known as a "query node".  Compare with findings node.

Test sensitivity:  The test sensitivity of an imperfect test is the percentage of cases testing positive out of those that should test positive.  See also test specificity.

Test specificity:  The test specificity of an imperfect test is the percentage of cases testing negative out of those that should test negative.  See also test sensitivity.

Text editor:  A text editor is a program that allows you to modify plain ASCII text.  For small files you can use Netica’s File New Text Edit or File Open as Text.  For larger files you can use a word processing program such as MS Word or WordPad (save as “Text Only” if you wish to make a file).

Text file:  A text file consists only of ASCII characters.  It has no special symbols, no formatting (bold, italics, fonts or sizes), no structure (paragraph sections, chapter sections, margins, etc.), and no special inserts (pictures, tables, etc.).  You can create or modify a text file using a text editor.

Time-delay link:  A time-delay link is a link which indicates that the child node is for a variable's value at a point in time later than the value of the parent node on which it depends.  The time difference is something that you can set for the link.  A net that contains time-delay links is not a normal Bayes net, but rather a meta-level representation, but it can be expanded to produce a normal Bayes net.  Time delay links can be used to model feedback.  more info

Toggled:  A menu item that can be toggled is like a switch that can be turned on and off.  When you choose it, a check-mark will appear beside it, indicating it is on.  Choosing it again will remove the check-mark, indicating it is off.

Triangulated:  An undirected network is triangulated if every loop of length 4 or more has a chord, i.e. a link joining two nonconsecutive nodes.  Triangulating a net is the process of adding links until it is triangulated (which may be done in different ways, depending on where the links are added).  The resulting net is composed only of triangles (each side being a link) fused together along their sides.

Undirected loop:  An undirected loop (sometimes just called a “loop”) is a path through a net, not necessarily following the direction of the arrows, which returns to its beginning (i.e. the first node of the path is the same as the last).  Compare with directed cycle.

Undirected network:  An undirected network is one where the links have no direction (i.e. no arrows).  A Markov network is an example.  Compare with a directed network.

Unicode:  Unicode is a text character set that was designed with the goal of being able to represent all the characters of all the world's languages.  It is by far the dominant international character set, and was developed by a consortium with most of the largest computer companies as members.  The first version was released in 1991, and every so often a new version is released having more characters.  It normally represents each character with 2 bytes, although there are a number of complexities and alternate encodings.  Netica allows most aspects of Bayes nets to be represented in Unicode, while the others are required to be in ASCII.

User reports: a Netica mechanism which displays customized information pertaining to a node, group of nodes, or to an entire net.  The user report could be as simple as a text message giving a more detailed description of what a node means.  Or it could be more complex, such as the current belief probabilities of the nodes, or a sensitivity analysis of the net. User reports can also be generated through the HED system.

Utility node:  A utility node (also known as a “value node”) is a node in a decision net whose expected value is to be maximized while searching for the best decision rule for each of the decision nodes.  It is usually drawn as a flattened hexagon or a diamond.

Weak link:  The absence of a link always indicates some independence, but even if a link is present there may nearly be independence if the relationship that the child node has with its parents indicates that.  Then it is called a weak link.  Removing a weak link has little effect on the full joint probability distribution, and therefore little effect on inference results.

Whitespace:  Whitespace is text which is composed solely of spaces, tabs, new lines, carriage returns or other ascii control characters.

 

Home > Reference > Glossary > Glossary S to Z