Generative quantum learning of joint probability distribution functions

Modeling joint probability distributions is an important task in a wide variety of fields. One popular technique for this employs a family of multivariate distributions with uniform marginals called copulas. While the theory of modeling joint distributions via copulas is well understood, it gets pra...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:Physical review research Ročník 4; číslo 4; s. 043092
Hlavní autoři: Zhu, Elton Yechao, Johri, Sonika, Bacon, Dave, Esencan, Mert, Kim, Jungsang, Muir, Mark, Murgai, Nikhil, Nguyen, Jason, Pisenti, Neal, Schouela, Adam, Sosnova, Ksenia, Wright, Ken
Médium: Journal Article
Jazyk:angličtina
Vydáno: American Physical Society 01.11.2022
ISSN:2643-1564, 2643-1564
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:Modeling joint probability distributions is an important task in a wide variety of fields. One popular technique for this employs a family of multivariate distributions with uniform marginals called copulas. While the theory of modeling joint distributions via copulas is well understood, it gets practically challenging to accurately model real data with many variables. In this paper, we show that any copula can be naturally mapped to a multipartite maximally entangled state. Thus, the task of learning joint probability distributions becomes the task of learning maximally entangled states. We prove that a variational ansatz we christen as a “qopula” based on this insight leads to an exponential advantage over classical methods of learning some joint distributions. As an application, we train a quantum generative adversarial network (QGAN) and a quantum circuit Born machine (QCBM) using this variational ansatz to generate samples from joint distributions of two variables in historical data from the stock market. We demonstrate our generative learning algorithms on trapped ion quantum computers from IonQ for up to eight qubits. Our experimental results show interesting findings such as the resilience against noise, outperformance against equivalent classical models and 20–1000 times less iterations required to converge as compared to equivalent classical models.
ISSN:2643-1564
2643-1564
DOI:10.1103/PhysRevResearch.4.043092