Sample Beans
The following examples are packaged with the ABLE code. Each
example has a corresponding directory in the /able/examples
subdirectory. In the directory you will find the source code to
the example as well as related data or rule files.
Files in the examples/ableagent
directory show how to create a custom AbleAgent by extending the
AbleDefaultAgent class, complete with source, BeanInfo file, and
an example Customizer.
Files beginning with SimpleAbleBean* in the examples/ablebean directory show how to
easily create a custom AbleBean by extending the AbleObject class,
complete with source, BeanInfo file, and an example Customizer.
Files beginning with AbleFileWatcher* in the examples/ablebean directory are referred
to in Extending Able.
Files in the examples/fuzzy directory
show how to use Sensors and Effectors and contain various example
rule sets.
There are several GeneticObject examples in the examples/genetic directory. To access
these, create a GeneticSearchAgent and specify the class function
in the bean customizer dialog.
The following neural network examples can be found in the examples/neural directory:
- The animal definition
file can be used to create a neural classification agent.
- The XOR definition file
can be used in a back propagation network to learn to
predict a binary function.
- CoinIdentifier contains data describing common United
States coins. It can be used as a neural classification
agent. After the network is trained, you can edit the
Import bean generated to interactively describe the coins
in your pocket and see if the result is correct.
- CreditScore contains both train and test data to
calculate the credit limit granted to applicants. It can
be used in a neural classification agent. Custom filters
are needed for data representation so that the network
will train properly.
- MarketAnalysis is used to create clusters of customer
profiles from customer purchasing history. It can be used
in a neural clustering agent.
- SalesForecast contains both train and test data used to
predict the percentage change in gross retail sales from
one day to the next. It can be used in a neural
prediction agent.
- Tdwalk contains the data used to train a temporal
difference learning network. Each input value represents
a position, and each record indicates an object's
position is at a point in time. The pattern ends when the
object has reached the leftmost or rightmost position,
and network learns to predict whether the object is going
to the right or the left.
Note that the suggested use for each example is not
necessarily the only model that can be used with the data
provided.