Click on a network's title to view a diagram of the network. You can
also click on "(DNET File)" to download the Netica-readable
version of the network (a free
demo program is available to work with the downloaded network).
Each network is credited to the source from which it was obtained.
If it originally comes from some other source please let us know.
| Alarm.....(DNET file) |
| ALARM stands for 'A Logical Alarm Reduction Mechanism'. This is
a medical diagnostic system for patient monitoring. It is a nontrivial
belief network with 8 diagnoses, 16 findings and 13 intermediate
variables. Described in BeinlichSCC89. |
| |
| Asia.....(DNET
file) |
| A very small belief network for a fictitious medical example about
whether a patient has tuberculosis, lung cancer or bronchitis, related
to their X-ray, dyspnea, visit-to-Asia and smoking status. Also
called "Chest Clinic". From LauritzenSpiegelhalter88. |
| |
| Boerlage92.....(DNET
file) |
A subjective belief network for a particular scenerio of neighborhood
events, that shows how even distant concepts have some connection.
From Boerlage92. |
| |
| Bouncing.....(DNET
file) |
| A time delay belief network for the position and velocity of a
frictionless 'ball' bouncing between 2 barriers. From Norsys. |
| |
| Cancer.....(DNET
file) |
| A very small belief network for a fictitious medical example about
whether a patient has metastatic brain cancer based on headaches,
coma and serum calcium. Originally from Cooper84 (PhD thesis), it
has appeared in Spiegelhalter86,
Pearl88 (p. 196), and
Neapolitan90 (p.
179). The node names are chosen to match historical choices. Neapolitan
uses different conditional probabilities; see the file Cancer_Neapolitan.
|
| |
|
Car Buyer.....
(DNET File) |
| An example influence diagram for Joe, who has to decide whether
to buy a certain used car which may be a 'peach' or a 'lemon'. He
has the option of doing some tests beforehand, and of buying it
with a guarantee or not. This is the classic example of an influence
diagram derived from a decision problem with a very asymmetric decision
tree, since if Joe decides not to test the car, then the test results
have no meaning, etc. This problem was posed (in decision tree representation)
by Howard62, and is described
as an influence diagram in Qi94 and in SmithHM93.
|
| |
| Car_Diagnosis_2
.....(DNET
file) |
| A simple example belief network for diagnosing why a car won't
start, based on spark plugs, headlights, main fuse, etc. This example
is small and just calls out to be expanded. From Norsys. |
| |
| DNET
Negative Suite |
| Archive of files to test software which reads DNET files. This
is a collection of files which violate increasingly more complex
aspects of the format. Good software reading the files should generate
an appropriate error message for each. |
| |
| DNET
Positive Suite |
| Archive of files to test software which reads DNET files. This
is a collection of files which should all read correctly, and which
test increasingly more complex aspects of the format. |
| |
| False_Barrier.....(DNET
file) |
| Evidence at node A creates a change of belief at node D, without
changing beliefs at B or C, even though B and C d-separate A from
D (this is possible because the joint beliefs of B and C change).
From Boerlage92. |
| |
| Fire.....(DNET
file) |
| A very simple belief network used by David Poole to introduce
belief networks in the early days. Somebody reports people leaving
a building because the fire alarm went off, but is it because of
tampering or is there really a fire (which would be the case if
smoke was observed)? From PooleNeufeld88.
|
| |
| Inverted_Pendulum.....(DNET
file) |
| A time delay decision network for the control problem known as
"inverted pendulum", or "pole balancing" in which the bottom of
a pole is moved around (in 1 dimension for this network) so that
the pole doesn't tip over. Problem described in MillerSW90.
Decision network from Norsys. |
| |
| Matheson90_Space_Mission.....(DNET
file) |
| A classic influence diagram used as an example of how to do value-of-information
calculations. A decision is to be made whether to send a space mission
to Mars or Venus, where the success of the mission is uncertain.
Among other things, we find the value of perfect information on
the reliabilities of the launch subsystems and the descender subsystems,
which can then be used to decide whether to do further testing.
From Matheson90.
|
| |
| Oil_Wildcatter |
| A classic influence diagram with decisions of whether to do seismic
tests for oil, and whether to drill for oil, in order to maximize
profits. In wide usage (e.g. Ross Shachter example), but originally
from Raiffa68.
Currently not included in distribution. |
| |
| Oil_Wildcatter_Extended....(DNET
file) |
| A decision network based on the "Oil Wildcatter" influence diagram
of Raiffa68,
but this network includes drilling for gas, has multiple value nodes,
and eliminates some no-forgetting links. This network contains only
dependence information; it has no numerical probabilities. From
Zhang93,
p. 21. |
| |
| Oil_Wildcatter_Simplified.....(DNET
file) |
| An influence diagram with decisions of whether to do seismic tests
for oil, and whether to drill for oil, in order to maximize profits.
Same as Oil_Wildcatter, but with some nodes absorbed ("summed out").
In wide usage, but originally from Raiffa68.
|
| |
| Thermostat_A.....(DNET
file) |
| A time delay decision network for the thermostat-heater
control problem. This is a simple example of heater control, with
a single heater, single thermal mass, single sensor, and costs for
overheating, underheating, energy and switching the heater on and
off. It could easily be expanded into a more complex example. From
Norsys. |
| |
| Umbrella.....(DNET
file) |
| A basic 4 node influence diagram for tutorial purposes. Makes
the decision of whether to take an umbrella given the weather forecast.
From Ross Shachter. |
| |
These nets are provided to be of use when studying the indicated books or papers;
otherwise, they may not be understandable.
| Cancer_Neapolitan.....(DNET
file) |
| This network is the same as the classic "Cancer" network, but
is the Neapolitan version (from Neapolitan90,
p. 179), which uses different conditional probabilities. |
| |
| Car
Buyer_Neapolitan.....(DNET
file) |
| A simplified version of the classic Car_Buyer influence diagram.
From Neapolitan90,
page 380. |
| |
| Neapolitan90_p261.....(DNET
file) |
| From Neapolitan90,
where it started as problem 5.5.2, p.183, it becomes example 7.5,
p. 261 (with diagram on p. 259), and continues numerically on p.
279. Originally based on LauritzenSpiegelhalter88. |
| |
| Neapolitan90_p386.....(DNET
file) |
| From Neapolitan90,
page 386. |
| |
| Puterman90_p337.....(DNET
file) |
| From Puterman90,
page 337. |
| |
| Separable1.....(DNET
file) |
| The simplest example of a separable (termed 'abnormal' by Zhang) decision net, and
the 2 nets it can be separated into. These networks contain only
dependence information; they have no numerical probabilities. From
Norsys. |
| |
| Separable2.....(DNET
file) |
| A simple example of a separable (termed 'abnormal' by Zhang) decision net, and
the 2 nets it can be separated into. See Separable1 for an even
simpler example. These networks contain only dependence information;
they have no numerical probabilities. From Norsys. |
| |
| SpiegelhalterDLC93.....(DNET
file) |
| A 20 node example of a belief net for medical diagnosis. This
file does not contain the numerical probabilities (except those
few given in the paper). This, together with the paper, provide
a good worked out example for clique tree (i.e. join tree) compiling
and propagation. From SpiegelhalterDLC93.
|
| |
| Zhang93_p121.....(DNET
file) |
| Example decision network to demonstrate 'removable' links and
nodes, which are a generalization of the barren nodes in influence
diagrams. This network contains only dependence information; it
has no numerical probabilities. From Zhang93,
page 121. |
| |
| Zhang93_p128.....(DNET
file) |
| Example of a separable (termed 'abnormal' by Zhang) decision network.
Also shows the 2 networks it can be separated into. These networks
contains only dependence information; they have no numerical probabilities.
From Zhang93, page 128. |
| |
| Zhang93_p55.....(DNET
file) |
| The 'Two Spy' example of a non-SDDN decision network. Policies
must be found for 2 agents who will then go into enemy territory
and not be able to communicate with each other. Based on the type
of enemy movement they observe, each must decide whether to send
a report back to home base, knowing they may be dangerously exposed
if they do, and that the report might be useless if the other agent
has already reported. This network contains only dependence information;
it has no numerical probabilities. From Zhang93,
page 55. |
| |
Bayes Net Library