Separate performances

When any function f (x ₁ , x ₂ ) assumed to depend on two generic coordinates x ₁ and x ₂ is known, one way of constructing its separate representation is to build the matrix f _j,k (x _1,j , x _2,k ), j = 1,..., J and k = 1,....K, to apply the SVD (singular value decomposition) and, as explained above for the POD, retain only the modes associated with the most significant eigenvalues $\left({\mathrm{\lambda }}_r\ge \epsilon {\mathrm{\hspace{0.17em}\lambda }}_1\right)$...