Tensor Decomposition Algorithms

Implementation and comparison of three CP (CANDECOMP/PARAFAC) tensor decomposition algorithms: Jennrich's algorithm, CP-ALS, and Ortho-ALS.

Algorithms

• Jennrich's Algorithm: Spectral method using generalized eigenvalue decomposition
• CP-ALS: Standard Alternating Least Squares
• Ortho-ALS: ALS with QR orthogonalization after each update

Usage

See run.ipynb for example usage.

Experiments

• Experiment 1: Bad conditioning test (varying rho)
• Experiment 2: Noise sensitivity (varying sigma)
• Experiment 3: Initialization sensitivity

Run with experiment1(), experiment2(), experiment3() from experiments.py.