Implementations and Demonstrations

Pei Wang


Since January 2008, NARS has been developed as open source software project Open-NARS.

There are several versions of the system with working examples for download: To run such a Java version, double clicking the JAR file in it, or typing the following line in a command window:

    java -jar "NARS.jar"

To run such a Clojure version, use one of the start scripts as specified in the README file.

To understand how the program works, start with the GUI guide of version 1.5.x or version 1.6.x, as well as the guide on I/O format.

Working Examples:

Working example files are included in the download ZIP packages, and they are explained in the following wiki pages:

After trying them, you can modify them, as well as make your own examples.

Testing Reports:

The expressing power of Narsese and the inferential power of NARS have been tested in various domains:

It is a Prolog program that implements all the inference rules in NAL 1-6 (version 1.1, 2013), without memory structure and control mechanism for multiple-step inference. The source code comes with a user's guide and examples. The examples are tested with SWI-Prolog.

The following two Prolog programs show the potential of NARS in natural language processing, as described in a simple example. The first program carries out grammar rule learning, and the second program applies the rules to novel sentences.

NAL truth-value functions

All the NAL truth-value functions (as specified in the 2013 book) are collected in a spreadsheet file NAL-TruthFunctions.xls.

The file contains three tables showing the relations among uncertainty measurements, the truth-value functions of the one-premise rules, and the truth-value functions of the two-premise rules, respectively. In each table, the input values can be modified, and the outputs of the functions change accordingly. The system parameters used in the functions can also be adjusted.

Related Works