Tensor Notes 3: tc-SVD, a Tensor SVD with a Mirror Boundary

The t-SVD quietly assumes your data wraps around along its third mode. Swapping that for a mirror boundary gives the tc-product and tc-SVD: real arithmetic through the DCT and consistently shorter run times.

Tensor Notes 2: An SVD for Tensors, and the t-Product Behind It

Tensors have more than one SVD, and each one gives something up. A tour of Tucker, CP and the t-product behind t-SVD, with a short NumPy implementation you can run.

Tensor Notes 1: Stop Flattening Your Data

Most machine learning pipelines turn images and videos into long vectors and lose their structure on the way. A plain introduction to tensors: modes, fibers, slices, unfolding and the mode-n product.