IonQ has spent months partnering with GE Research to experiment with the use of quantum computing to model multi-variable distributions in risk management. The results are here and we are very proud of what GE Research has accomplished with us.
Together, we trained quantum circuits with real-world data on historical data indexes in order to predict future performance. This was done using hybrid quantum computing, in which some components of a problem are handled by a quantum computer while others are done by a classical computer. We did this by using copulas, which are mathematical tools for modeling joint probability distributions. We compared those results with classical copula modeling approaches and found that in some cases our quantum predictions outperformed them, with the potential to recognize higher probabilities of instability events occurring in many industries.
The work was done using IonQ Aria, our 20 #AQ trapped ion quantum computer, which is currently available to select customers in private beta and will launch on Microsoft Azure later this year. In separate work that we jointly conducted last year with a major financial services company, we were able to model stock prices using IonQ Harmony, our 11 qubit quantum computer currently available to customers on all three major cloud platforms. That experiment modeled the probability of black swan events using two variables. In this new work using IonQ Aria, we were able to increase that to three and four variables in different indexes.
Optimization is always a crucial part of hybrid quantum computing. In the process of conducting this work with GE Research, we created an innovative form of optimization called annealing training (read the paper for more details). We now have a set of techniques and tools that we can continue to scale up for models that may continue to outpace classical computing methods as we build even more powerful quantum computers.
Since the term “copula” was coined in 1959, copulas have been used on classical computers for risk management in finance, medicine and engineering. With the help of quantum entanglement, we can now generate correlations among different qubits that have no classical equivalent. We believe that this research into the use of quantum copulas also has the potential to be useful in manufacturing and supply chain management. This is a new way to weigh the likelihood of dramatic changes in the course of a long and robust set of data. We believe that these results confirm that quantum copulas can outperform predictions about large sets of data which are made using only classical computing hardware.
To find out how we can help your organization using our hardware and expertise, please contact us. Stay on top of news and research at IonQ by signing up for our email list.