Figure 2.

The optimization process of GP-MI for the DNILMF model parameters. At irritation t: Step 1: get $x_t$ according to the previous query ${\chi }_{t-1}$ and observations ${\mathcal{Y}}_{t-1}$; Step 2: if $x_t$=$x_{t-1}$ or t is equal to the max value, exit the program; if not, put $x_t$, the disease kernel matrix, lncRNA–disease association matrix, and lncRNA kernel matrix into the DNILMF model, and we can get output $x_t$. Then, take $x_t$ and $y_t$ as the start of the next irritation.