Table 4 Weights for each data sources computed by the RB-KNN method (glm.1111). The mean and standard deviations of weights of base similarity measures are presented based on the 5-old cross validation. For the ease of presentation, these weights are computed using the linear model, whose performance is slightly worse than the default quadratic model (accuracy of 0.65 as compared to 0.68).

From: A Regression-based K nearest neighbor algorithm for gene function prediction from heterogeneous data