Domain Adaptation with CPD Algorithm

[williamrxu]

Is your feature request related to a problem? Please describe.
This is my individual portion of the sprint 1 issue splitting from #79 domain adaptation. I am trying to use the CPD algorithm (https://arxiv.org/pdf/0905.2635.pdf) to domain adapt training data from one task to another task to provide twice the number of training data.

Additional context (e.g. screenshots)
Experiment:

• XOR and rotated XOR at different angles and evaluating Generalization Error