Periodic Reporting for period 1 - SINGULARITY (SINGULARITY: The first quantum software toolbox for finance)
Periodo di rendicontazione: 2022-09-01 al 2023-08-31
Quantum computers are reaching a scale where they can start delivering value to many industries. The quantum software market is estimated to grow at a CAGR of 42% over the next five years to reach €1.42 billion by 2027. McKinsey highlighted finance as the earliest high-impact application of quantum computing. We have been pioneers in this space. We authored one of the seminal papers in the field.
Then we developed a quantum software toolkit for finance, and results from beta testing with leading financial institutions brought us to the limelight. Now, thought leaders like McKinsey, The Economist, BCG, and Forbes point at Multiverse as a leading firm in quantum finance. Quantum computing is a strategic technology for Europe. Anglo-Saxon firms dominate both quantum and finance. Yet, Multiverse is leading this race.
SINGULARITY is the first quantum software toolkit for finance. The toolkit leverages the best of classical and quantum hardware to tackle financial problems in their full complexity. SINGULARITY contains several proprietary quantum and quantum-inspired algorithms based on quantum machine learning, tensor networks, and quantum annealing. These methods outperform state-of-the-art approaches in investment optimisation, capital allocation, and risk management. Their performance has been stringently tested in proof-of-concept trials with leading financial institutions using real quantum computers. While SINGULARITY owes its performance to advanced quantum techniques, its interface is intuitive and requires no specialised quantum computing knowledge. The toolbox does all the preprocessing needed to make financial data quantum-ready autonomously. Thus, SINGULARITY can be seamlessly integrated into standard financial software and used by quantitative staff without further training.
In this project we focus on developing the algorithms and tools to be integrated into the SINGULARITY financial toolkit. In particular, in this project we focus on developing tools that can be used to help improve the stability of the global financial markets by improving securities pricing and risk management, and global network analysis techniques .
Then we developed a quantum software toolkit for finance, and results from beta testing with leading financial institutions brought us to the limelight. Now, thought leaders like McKinsey, The Economist, BCG, and Forbes point at Multiverse as a leading firm in quantum finance. Quantum computing is a strategic technology for Europe. Anglo-Saxon firms dominate both quantum and finance. Yet, Multiverse is leading this race.
SINGULARITY is the first quantum software toolkit for finance. The toolkit leverages the best of classical and quantum hardware to tackle financial problems in their full complexity. SINGULARITY contains several proprietary quantum and quantum-inspired algorithms based on quantum machine learning, tensor networks, and quantum annealing. These methods outperform state-of-the-art approaches in investment optimisation, capital allocation, and risk management. Their performance has been stringently tested in proof-of-concept trials with leading financial institutions using real quantum computers. While SINGULARITY owes its performance to advanced quantum techniques, its interface is intuitive and requires no specialised quantum computing knowledge. The toolbox does all the preprocessing needed to make financial data quantum-ready autonomously. Thus, SINGULARITY can be seamlessly integrated into standard financial software and used by quantitative staff without further training.
In this project we focus on developing the algorithms and tools to be integrated into the SINGULARITY financial toolkit. In particular, in this project we focus on developing tools that can be used to help improve the stability of the global financial markets by improving securities pricing and risk management, and global network analysis techniques .
Throughout the current duration of the project we have developed quantum solutions for complex problems tackled in finance. Although in some applications quantum technologies may be limited by current hardware development we employ quantum inspired tensor network techniques to provide another set of unique advantages.
One of the most important computational problems in finance is the pricing of complex securities such as derivatives. Using pure quantum we were able to construct novel circuits to price derivatives using quantum Monte Carlo on the latest IBM QPU. Furthermore, we developed a new tensor network enhanced deep learning framework for pricing Bermudan style derivatives, we applied this methodology to Swaption contracts and were able to produce more accurate and reliable prices than compared to current industry standard Monte Carlo approaches.
Other work related to pricing focuses on the application of quantum Monte Carlo for determining the fair value of a portfolio of assets with an underlying stochastic model of growth. As an alternative to pure quantum we develop a tensor network integration framework that uses efficiently encoded probability distributions and evaluation functions .
With respect to macroeconomic stability in this project we have developed an improved quantum methodology for analysing the stability of large financial networks opening the doors to practically relevant applications.
The key aspect of the Singularity project is bringing quantum computing to the inexperienced or less technical user, making quantum computing advantages accessible to all. In order to achieve this we have been developing intuitive middleware software to connect users to our solutions, such as:
- an Application Programming Interface (API) accessible programmatically via Python;
- an AWS backend infrastructure that handles the API requests and runs the solutions on the required quantum or Graphics Processing Unit (GPU) hardware;
- user interfaces as excel plugins for our financial solutions, such as derivatives pricing.
One of the most important computational problems in finance is the pricing of complex securities such as derivatives. Using pure quantum we were able to construct novel circuits to price derivatives using quantum Monte Carlo on the latest IBM QPU. Furthermore, we developed a new tensor network enhanced deep learning framework for pricing Bermudan style derivatives, we applied this methodology to Swaption contracts and were able to produce more accurate and reliable prices than compared to current industry standard Monte Carlo approaches.
Other work related to pricing focuses on the application of quantum Monte Carlo for determining the fair value of a portfolio of assets with an underlying stochastic model of growth. As an alternative to pure quantum we develop a tensor network integration framework that uses efficiently encoded probability distributions and evaluation functions .
With respect to macroeconomic stability in this project we have developed an improved quantum methodology for analysing the stability of large financial networks opening the doors to practically relevant applications.
The key aspect of the Singularity project is bringing quantum computing to the inexperienced or less technical user, making quantum computing advantages accessible to all. In order to achieve this we have been developing intuitive middleware software to connect users to our solutions, such as:
- an Application Programming Interface (API) accessible programmatically via Python;
- an AWS backend infrastructure that handles the API requests and runs the solutions on the required quantum or Graphics Processing Unit (GPU) hardware;
- user interfaces as excel plugins for our financial solutions, such as derivatives pricing.
Whilst full benchmarking of algorithms is to be performed in up and coming tasks, initial studies have been able to demonstrate the utility of our approaches.
Our most successful results show how the combination of quantum inspired tensor neural networks and a new deep learning scheme for Bermudan options is able to provide more accurate prices than industry standard Monte Carlo. Monte Carlo approaches are well known to underestimate the price of Bermudan options and scale poorly with the complexity of the underlying model. Our approach resolves both these issues, the tensorized neural networks allow for significantly improved training speed and stability of deep learning. Bermudan style swaptions play an important part in the trillion dollar OTC derivatives market, consequently even small improvements in prices and risk analysis can improve operational efficiency and risk management having a large impact for financial institutions and the market.
For other financial problems we demonstrate using tensor network integration as an alternative to quantum Monte Carlo integration. Our initial results are encouraging and demonstrate the efficiency and accuracy of the method. Although practical quantum Monte Carlo techniques are still some time away, tensor network integration looks to be a powerful and accessible alternative. There are many computational problems where this technique can be utilised in finance.
With respect to macroeconomic analysis our key result in this project has been the ability to extend previous algorithms to practical sized applications by improving the scalability. Existing approaches for network stability on quantum computers required an exponential number of qubits to encode the necessary crash behaviour using approximations. Our quantum approach exactly encodes the behaviour and only scales at most quadratically, resulting in a more accurate and scalable algorithm. Using this approach we hope to demonstrate how real-world networks can be analysed giving deeper insight into our current global financial system and its stability.
Our most successful results show how the combination of quantum inspired tensor neural networks and a new deep learning scheme for Bermudan options is able to provide more accurate prices than industry standard Monte Carlo. Monte Carlo approaches are well known to underestimate the price of Bermudan options and scale poorly with the complexity of the underlying model. Our approach resolves both these issues, the tensorized neural networks allow for significantly improved training speed and stability of deep learning. Bermudan style swaptions play an important part in the trillion dollar OTC derivatives market, consequently even small improvements in prices and risk analysis can improve operational efficiency and risk management having a large impact for financial institutions and the market.
For other financial problems we demonstrate using tensor network integration as an alternative to quantum Monte Carlo integration. Our initial results are encouraging and demonstrate the efficiency and accuracy of the method. Although practical quantum Monte Carlo techniques are still some time away, tensor network integration looks to be a powerful and accessible alternative. There are many computational problems where this technique can be utilised in finance.
With respect to macroeconomic analysis our key result in this project has been the ability to extend previous algorithms to practical sized applications by improving the scalability. Existing approaches for network stability on quantum computers required an exponential number of qubits to encode the necessary crash behaviour using approximations. Our quantum approach exactly encodes the behaviour and only scales at most quadratically, resulting in a more accurate and scalable algorithm. Using this approach we hope to demonstrate how real-world networks can be analysed giving deeper insight into our current global financial system and its stability.