Research On Algorithmic Relationship Between Programming Resources On The Web

ACM International Collegiate Programming Contest (ACM) is known as the largest scale and the highest level. There are many Online Judge Systems (OJ) provide huge problems to students to help them improve their achievement on the web. Students submit their program source codes to the OJ to solve problems. Algorithm knowledge scatters in the programming resources which are similar to OJ, but it is not appeared in any structure. Intelligent Computer Assisted Instruction (ICAI) is a personalized OJ and can provide corresponding algorithm knowledge to different students. It is necessary to discovery the algorithmic relationships for programming resources on the web.Our research group has researched the language points and knowledge unit and all their objects are language points. At the same time, we have researched algorithm recognition and algorithm knowledge in the problem-solving reports. The work of algorithm recognition hasn’t given the suitable representations for algorithms. And the research on problem-solving reports is based on fixed algorithm knowledge and hasn’t discovery the algorithm knowledge in program source codes.Our work is to discovery the structure of algorithm knowledge. There are two kinds of algorithmic relationships between programming resources on the web:structural similarity relationship and associative relationship. Firstly, we convert all program source codes into vectors and they are representations for algorithm knowledge of source codes. Because Lobe Component Analysis (LCA) can deal with samples which are muddy, we use one-layer neural network based on LCA to develop algorithms from vectors. We choose adult algorithms from the neural network which are developed. We can get the relationship between algorithms and problems through the adult algorithms. Secondly, we obtain the algorithmic signatures from the matrixes of developed algorithms and get the algorithmic descriptions through analyzing program source codes belong to .each algorithm. Finally, algorithm knowledge is organized by hierarchical model according to the algorithmic relationships between programming resources on the web and we present the prototype system.This paper organizes the algorithm knowledge based on hierarchy. The algorithmic relationships benefit the personalized programming tutoring.