AI-MachineLearning deletionCandidateAlgorithm

Basic-Logic：

The Deletion Candidate Algorithm is used to collect and summarize the characteristics of the samples to obtain the variant space, and then the unknown samples can be classified. It is based on the analysis and summary of the sample attribute set. The basic principle is as follows:

1. Set up two hypotheses, one is a general hypothesis, the other is a special hypothesis. Through the analysis of each sample, expand the special hypothesis and narrow the general hypothesis.

2. The hypothesis is the setting of each attribute value of the sample, which can be set to a fixed value or an unknown value. That is, any value can be accepted, and if it is set to nothing, it means that no attribute value is accepted. Right

3, for each line of analysis, extract the various attributes of the sample of each line

1) For positive case:

It is necessary to delete the hypothesis that does not agree with the positive example in the general hypothesis

Attributes in the special hypothesis that do not agree with the positive example need to be removed (that is, changed to "? "). Start with the special assumption that all is NOTHING, that is, no attribute value is accepted

2) For negative case:

You need to delete the assumptions in the general hypothesis that are consistent with the counterexample, and then add the attribute values that are inconsistent with the counterexample, and the other attribute values are "?" Assumption of

3) The essence is that the general hypothesis illustrates the characteristics of the counterexample, and the special hypothesis illustrates the characteristics of the positive example

example1：

Learning the concept of  ' circle ' with the Deletion Candidate Algorithm

size: lg sm  shape: cir squ tri

step1: G={(x,y)}

           S={(sm,cir),(sm,squ),(sm,tri),(lg,cir),(lg,squ),(lg,tri)}

step2: select (sm,cir) as a positive case  //default the first element in S as the positive case

        G={(x,y)}  

        S={(sm,cir)}  //delete the negative case from G, then ,select the element cover the positive example

step 3:  (lg,squ) negative case

        S={(sm,cir)}

        G={(sm,y),(x,cir)}

step4: (lg,cir) positive case

        G={(x,cir)}

        S={(x,cir)}

G=S