Inducing knowledge from examples
The paper presents way of explaining knowledge induction process
to help students to understand it, and experiment by means of computer
program. The knowledge is induced from examples of some situation with
a sort of generalization that use entropy measure to guide
the generalization process. Set of examples can be changed to enable
analysis of influence of new examples to the induced knowledge. Simple
addition of a new example that describe the same situation, possibly leads
to modification of the induced knowledge.
The process is illustrated with the simple algorithm
implemented in Logo. Furthermore, the algorithm is explained
in details to enable simple modification and use with some other
input examples. The induction algorithm implemented in Logo
is one of the well known machine learning algorithms. It has been developed
in a few ways to some poverful algorithms that have been successfully
applied on a lot of real-world problems, to induce knowledge about
some real situation. One of widely known real-world application
is in medical diagnosis where machine learning system is used to predict
a patient diagnosis from the symptoms.