Researchers have successfully demonstrated a quantum algorithm for Monte Carlo simulations on Amazon Braket, a cloud-based quantum computing platform. The team used PennyLane’s QuantumMonteCarlo module to simulate the pricing of an Asian call option, a complex financial derivative. The results show a quadratic advantage in estimation error scaling over classical Monte Carlo methods, confirming theoretical predictions. This breakthrough has significant implications for the finance industry, where Monte Carlo simulations are widely used to estimate the prices of complex derivatives.
The study builds on previous work by Ashley Montanaro, who proposed the quantum speedup of Monte Carlo methods. The team also utilized Amazon Braket’s capabilities, such as Hybrid Jobs and GPU instances, to push the limits of their simulations. Key individuals involved in this research include Alexander Dalzel from the AWS Center for Quantum Computing and Patrick Rebentrost, a leading researcher in quantum computational finance.
The authors have demonstrated a quadratic speedup of quantum Monte Carlo (QMC) over classical Monte Carlo methods in estimating the price of an Asian Call option. The QMC algorithm achieves this by varying the discretization qubits (nd) and phase estimation qubits (npe).
The authors show the scaling of error estimation in both classical and quantum algorithms for different values of nd, with a fixed maturity time of one year and two time-intervals for averaging. The results indicate that the exact scaling depends on various factors, but it remains roughly quadratic.
The authors plot the ratio of quantum to classical scaling exponents (zQ /zc) for a range of values of nd. The results show an almost quadratic speedup for larger values of nd, while smaller values result in a less-than-quadratic speedup. This is expected, as lower values of nd can cause higher errors due to discretization, which may impact the overall speedup of QMC.
They also compare the runtime of the simulations performed here, plotting the computational wall-time against varying phase estimation qubits (npe) for five different values of discretization qubits (nd). The results show that the wall-time grows exponentially with increasing npe for each nd. Additionally, for a fixed npe, a higher nd results in longer simulation wall-time.
The conclusion is clear: QMC algorithms can be used to estimate the price of Asian Call options with a quadratic advantage over classical Monte Carlo methods. This approach can be applied to other derivative financial instruments that are usually estimated using classical Monte Carlo methods.
They also have utilized PennyLane’s QuantumMonteCarlo module to simulate the QMC algorithm on PennyLane’s lightning.qubit simulator, and Amazon Braket provides capabilities to further push the limits of these simulations, such as parallelizing PennyLane simulations using Hybrid Jobs, accommodating higher values of nd and npe using larger AWS computational instances, and speeding up each simulation using GPU instances with lightning.gpu.
Overall, this study demonstrates the potential of QMC algorithms in financial derivative pricing, and highlights the importance of optimizing algorithm parameters to achieve a quadratic speedup over classical methods.
External Link: Click Here For More
