Table 1.
Pseudo-code for the K-OPLS model training algorithm. K denotes the original kernel matrix, Ki the kernel matrix deflated by i Y-orthogonal components and Qi the Ki matrix deflated by A predictive components.
| Step | Description | |
| 1. | Estimate the predictive Y-weights (Cp) by eigen-vector decomposition of YTKY | |
| 2. | Project Y onto Cp to achieve the predictive score matrix of Y: Up ← YCp | |
| 3. | Calculate the predictive score matrix of X: Tp ← KUp | |
| 4. | Repeat for i : 1 to Ao | |
| 4.1 | Estimate the Y-orthogonal loadings co by eigen-vector decomposition of TpTQiTp. | |
| 4.2. | Calculate the Y-orthogonal score vector: to,i ← QiTpco | |
| 4.3. | Deflate Ki by to,i, yielding Ki+1 | |
| 4.4. | Update the predictive score matrix: Tp ← Ki+1Up | |
| 5. | Predictions of Y: Yhat ← T* p (TpTTp)-1TpTUpCpT. For predictions of future samples, T* p originates from the prediction set. | |