qfeval_functions.functions.pca

pca(x)[source]

Computes principal components on the given input x. The returned value represents principal components. Specifically, result[*, i, :] represents the \((i+1)\)-th largest principal component of the batch specified by *. Shape:

  • x: \((*, S, D)\) where * means any number of additional dimensions, S means the number of sections, and D means the number of dimensions.

In qfeval, dimensions and sections often represent symbols and timestamps respectively. :returns:

components[i] represents the eigenvalue of the

\(i\)-th component.

  • explained_variance (Tensor): explained_variance[i, j] represents

    the \(i\)-th component’s weight for the j-th feature.

Return type:

PcaResult