Proofs and additional experiments on Second order techniques for learning time-series with structural breaks

This work provides theoretical and empirical support for an existing method for learning time-series with structural breaks, making it an incremental contribution to the field.

This paper provides complete proofs for the lemmas regarding the regularized loss function used in second-order techniques for time-series with structural breaks, originally proposed by Osogami (2021). It also presents experimental results that validate these techniques.