Machine learning via relativity-inspired quantum dynamics

Efficient estimation of trainability for variational quantum circuits

“…The effect of the structure of the cost function on the appearance of barren plateaus was also investigated in other works [14,15], and it was shown that global cost functions are more prone to exhibit barren plateaus. Note that shallow models such as quantum kernel machines [36][37][38][39][40] and reservoir computing models [41][42][43][44], while often easier to train than variational quantum algorithms, might also suffer from trainability issues of a similar nature [45].…”