Elicitation of machine learning to human learning from iterative error correcting

TLDR: This paper presents two iteratively error correcting based probabilistic neural networks (PNN) for connecting human learning and machine learning and proposes a recommendation approach of learning samples which selects samples according to density of knowledge points through calculating data field ofknowledge points covered by problems.

Abstract: Numerous high performance machine learning algorithms are designed based on human learning, while human learning can also acquire elicitation from machine learning to investigate highly efficient learning process. This paper presents two iteratively error correcting based probabilistic neural networks (PNN) for connecting human learning and machine learning. C-PNN, G-PNN and G-PNN have been used to delete redundancy samples in our learning software based on question bank. In detail, we propose a recommendation approach of learning samples which selects samples according to density of knowledge points through calculating data field of knowledge points covered by problems. The approach also deletes redundant problems in order to deal with the question-sea tactical and remedy the defects of random selecting usually used in human learning.