Research On Background Base Algorithms Based On Factor Space Theory

At present,artificial intelligence technology has made many amazing achievements in application,but its basic theory is still actively explored.Factor space theory is regarded as the mathematical basic theory of artificial intelligence.Based on the background basis theory of factor space,this paper originally proposes an improved background basis extraction algorithm(IBBE);designs a base point classification algorithm;proposes principal reasoning and studies principal reasoning solution.To the first problem,IBBE algorithm is proposed in the perception of algorithm expanding ability and diminishing datas noise ability.Differing from the previous ways,this new method uses—from inside to outside—order to extract base points,which is in line with knowledge expanding of the human.Meanwhile,noise datas can be filtered well by limiting steps' length.Futhermore,the concept of 2-Flou number is defined to structure ?-background base,because ?-base points can meet varous of decision maker and improve the generization ability of basepoints.Last,the method's rationality and availability is manifested by a demonstration.To the second problem,firstly,a kind of linear discriminated base-points classification(BPC)algorithm based on IBBE method is designed.Combining with the thought of kernel function,a nonlinear classifer is developed,named it with KBPC.Then,both BPC and KBPC show good performance in some typical data sets and even they can recognize new label,in this perspective,they have a kind of intelligent growth mechanism to some degree.To the last question,the formal definition of principal reasoning and related mathematical properties are put forward firstly.Then both the concepts of background distribution separation and base point distribution separation are defined,too.For solving problems with principal reasoning,combining the "divide and solve" method,the overall problem is divided into four sub-problems firstly,And then the algorithms and scientific computing are used to solve them one by one.Finally,accurate results are obtained.An example verifies the feasibility of each subalgorithm and the rationality of the overall design.