Affiliation:
1. Dipartimento di Fisica Università di Pavia Pavia Italy
2. Dipartimento di Ingegneria dell'Informazione Elettronica e Telecomunicazioni Università degli Studi di Roma “La Sapienza” Roma Italy
Abstract
AbstractIn classical machine learning (ML), “overfitting” is the phenomenon occurring when a given model learns the training data excessively well, and it thus performs poorly on unseen data. A commonly employed technique in ML is the so called “dropout,” which prevents computational units from becoming too specialized, hence reducing the risk of overfitting. With the advent of quantum neural networks (QNNs) as learning models, overfitting might soon become an issue, owing to the increasing depth of quantum circuits as well as multiple embedding of classical features, which are employed to give the computational nonlinearity. Here, a generalized approach is presented to apply the dropout technique in QNN models, defining and analyzing different quantum dropout strategies to avoid overfitting and achieve a high level of generalization. This study allows to envision the power of quantum dropout in enabling generalization, providing useful guidelines on determining the maximal dropout probability for a given model, based on overparametrization theory. It also highlights how quantum dropout does not impact the features of the QNN models, such as expressibility and entanglement. All these conclusions are supported by extensive numerical simulations and may pave the way to efficiently employing deep quantum machine learning (QML) models based on state‐of‐the‐art QNNs.
Funder
Ministero dell'Università e della Ricerca
Subject
Electrical and Electronic Engineering,Computational Theory and Mathematics,Condensed Matter Physics,Mathematical Physics,Nuclear and High Energy Physics,Electronic, Optical and Magnetic Materials,Statistical and Nonlinear Physics
Reference58 articles.
1. T.Condie P.Mineiro N.Polyzotis M.Weimer in2013 IEEE 29th Int. Conf. on Data Engineering (ICDE) IEEE Piscataway NJ2013 pp.1242–1244.
2. D.Dhall R.Kaur M.Juneja inProc. of ICRIC 2019 Springer NY2020 p.47.
3. A.Singh N.Thakur A.Sharma in2016 3rd Int. Conf. on Computing for Sustainable Global Development (INDIACom) Bharati Vidyapeeth's Institute of Computer Applications and Management New Delhi2016 pp.1310–1315.
4. The Problem of Overfitting
5. X.Ying inJournal of Physics: Conference Series Vol.1168 IOP Publishing Bristol UK2019 p.022022.