Elies Gil-Fuster
(Freie University Berlin)
 

The quest for successful variational quantum machine learning (QML) relies on the design of suitable parametrized quantum circuits, as analogues to neural networks in classical machine learning. To date, a few general design principles are known to ensure the resulting QML models display important good properties. Recent works have highlighted an intricate interplay between some of these properties, hinting at a fundamental obstacle for trainable QML models to yield a quantum advantage, which is still unresolved. In this talk we contribute to this debate. We give and motivate precise definitions, and we study the relation between trainability and dequantization of variational QML models. Our work offers ways forward toward the quest for good QML models.
The talk will be self-contained and broadly accessible.
 

Zoom registration link:
https://zoom.us/meeting/register/fl9NW1UnSsulOTxDnBICaw