Package lnsc

The LNSC (Laboratory for Natural and Simulated Cognition) package, contains the lab neural network simulator.

See:
Description

Interface Summary

FunctionalUnit	Interface supported by every objects that can be used like a function.
FunctionalUnitFactory	General interfaced shared by FunctionalUnit factories.

Class Summary

AbstractFunctionalUnit	Abstract class containing the basic implementation for the FunctionalUnit interface.
AbstractFunctionalUnitFactory	Basic internal functionalities for FunctionalUnit factories based on units prototypes.
AbstractSimpleUnit	Abstract class containing the basic implementation for simple univariate single-real-valued FunctionalUnit.
DataNames	This class contains constants for reserved keywords used by the package to store and retrieve data in a DataSet or in a DataSetCollection.
DataSet	Container for input patterns set and all relevant information such as output patterns set, target patterns set, error patterns set, internal values and others.
DataSetCollection	Zero-based ordered collection of data sets containing similar data.
FunctionalUnit.ProcessPatternResult	Return type for the method FunctionalUnit.processPattern(double[], boolean, boolean).
LinearAlgebra	Set of linear algebra functions for vectors and matrices.
LinearUnit	A univariate single-valued linear function.
LogisticUnit	A sigmoidal logistic unit.
Tools	Set tool functions.

Exception Summary

DataSetException	Exception raised when there are problems with a DataSet.
InvalidDataException	Exception raised when data under DataName is not of the appropriate type in a DataSet.
MissingDataException	Exception raised when a necessary DataName is missing from a DataSet.

Package lnsc Description

The LNSC (Laboratory for Natural and Simulated Cognition) package, contains the lab neural network simulator. This extremely flexible and extensible package includes numerous variations of Cascade-Correlation neural networks (including thh Knowledge-Based version) and Backpropagation neural networks. It also contains an implementation of KBANN as well as Optimal Brain Damage algorithm for cascaded networks. It contains many type of nodes and can easily be extended to use user-defined type of nodes. Similarly it contains many optimization algorithms and can easily be extended to user defined-ones. It also uses user-defined stopping criteria and can compute at run-time any user-defined statistics about the network under training on training and testing data or undergo any other user-defined analysis operation. Finally, all these, including the networks themselves can be recorded and saved for further analysis.

The library spirit is to make every single piece of a learning system a block and to build learning algorithm from these blocks. The building blocks of the library stand in one class and few interfaces.

• The class DataSet is a general data holder. It is used to put training or testing data and to gather outputs as well as intermediate results and calculations.
• The interface FunctionalUnit provides the general interface for any classifier or function approximator. A functional unit is a real-vector-valued multi-variate function that may or may not be differentiable. For example, single processing element in neural networks as well as whole networks are functional units. (The library contains around 17 functional units.)
• The page subpackage provide the necessary interface to build artificial agent moving in artificial world like game playing program, artificial grid-world, etc...
• Others subpackages are not necessarly part of the LNSC package.

This is actually only a subset of the LNSC package.

The code of this library is the property of Thomas R. Shultz and Francois Rivest. The conceptual structure of this library is the intellectual property of Francois Rivest. This package was developed for research purpose and any commercial use, and non-educational use is strictly prohibited unless explicit written consent is given by the author Francois Rivest. Copyright 1997-2008.