Research on Unbiased Recommendation Algorithm for Logical Neural Inference
During the recommendation process,logical neural reasoning demonstrated strong performance.However,existing methods introduce additional information about users and projects when recommending.This article proposes an unbiased personalized recommendation algorithm LNR-UR for logical reasoning,this method can construct inference processes through simple logical operations and construct neural inference models by determining the similarity of vectors in the logical space.Without adding additional information,extracting user preferences to reduce prediction bias through negative feedback sampling based on cross paired sorting.Experiments have shown that,compared with classical methods,the proposed algorithm has higher recommendation accuracy.
NETL
NSTL
万方数据
