Center for Analysis and Design of Intelligent Agents

Xiang Li's slides from the NARS tutorial given in 2018.

nars-tutorial.pdf

Grading: 10% of final grade.

final_project_fuzzy_logic.pdf

Fuzzy logic is a form of many-valued logic in which the truth values may be any real number between 0 and 1. It is employed to handle concepts of graded membership and proposition of partial truth with completely true and completely false.

NAL, the representation language of NARS, is not a “fuzzy logic” in the conventional sense, but shares certain intuitions and ideas with it. In NAL, fuzziness typically comes from the diversity in the intension of a term, that is, an instance may have some, but not all, properties that are usually associated with the concept. For example, “Penguins are birds“ is true to a degree, because penguins do not have all the common properties of birds, but only some of them. This distinguishes it from the other common forms of uncertainty, randomness, as randomness typically comes from the diversity in the extension. For example, “birds fly” is true to a degree, because some birds fly, and some do not. These two types of uncertainty are uniformly represented and processed in NAL, as different ways for a categorical statement to get (positive and negative) evidence.

©K.R.Thórisson
EOF