1.

Estimate the predictive Yweights (C_{p}) by eigenvector decomposition of Y^{T}KY

2.

Project Y onto C_{p} to achieve the predictive score matrix of Y: U_{p} ← YC_{p}

3.

Calculate the predictive score matrix of X: T_{p} ← KU_{p}

4.

Repeat for i : 1 to A_{o}


4.1

Estimate the Yorthogonal loadings c_{o} by eigenvector decomposition of T_{p}^{T}Q_{i}T_{p}.


4.2.

Calculate the Yorthogonal score vector: t_{o,i} ← Q_{i}T_{p}c_{o}


4.3.

Deflate K_{i} by t_{o,i}, yielding K_{i+1}


4.4.

Update the predictive score matrix: T_{p} ← K_{i+1}U_{p}

5.

Predictions of Y: Y_{hat} ← T* _{p} (T_{p}^{T}T_{p})^{1}T_{p}^{T}U_{p}C_{p}^{T}. For predictions of future samples, T* _{p} originates from the prediction set.
