Potential Quantum Computing Use Cases: Tutorials and algorithms
Currently, quantum computing is suitable for certain algorithms such as optimization, machine learning or simulation. With this type of algorithms, several use cases can be applied in different fields: financial services such as portfolio risk optimization, fraud detection, healthcare (drug research, protein study, etc.), supply chains and logistics, chemicals and petroleum or research for new materials are all areas that will be primarily impacted. Let’s explore in this article a couple of them.
Financial Services
Financial institutions has the potential to take advantage of quantum computing in many areas such as trading strategies improvement, improve management of client portfolios and better analyze financial risks. A quantum algorithm in development, for example, could potentially provide quadratic acceleration when using derivative pricing — a complex financial instrument that requires 10,000 simulations to be valued on a conventional computer, would only require 100 quantum operations on a quantum device. One of the use cases is the optimization of trading. It will also be possible for banks to accelerate portfolio optimizations such as Monte Carlo simulations. The simulation of buying and selling of products (trading) such as derivatives can be improved thanks to quantum computing. The complexity of trading activities in the financial markets is skyrocketing. Investment managers struggle to integrate real constraints, such as market volatility and changes in client life events, into portfolio optimization. Currently, the rebalancing of investment portfolios that follow market movements is strongly impacted by calculation constraints and transaction costs. Quantum technology could help reduce the complexity of today’s business environments. The combinatorial optimization capabilities of quantum computing can enable investment managers to improve portfolio diversification, rebalance portfolio investments to more precisely respond to market conditions and investor objectives, and to streamline more cost-effective transaction settlement processes. Machine learning is also used for a portfolio optimization and scenario simulation. Banks and financial institutions such as hedge funds are increasingly interested because they see it as a way to minimize risks while maximizing gains with dynamic products that adapt according to new simulated data. Personalized finance is also an area explored. Customers demand personalized products and services that quickly anticipate changing needs and behaviors. There are small and medium-sized financial institutions that can lose customers because of offers that do not favor the customer experience. It is difficult to create analytical models using behavioral data fast enough and precisely to target and predict the products that some customers need in near real time. A similar problem exists in detecting fraud to find patterns of unusual behavior. Financial institutions are estimated to lose between $10 billion and $ 40 billion in revenue annually due to fraud and poor data management practices. For customer targeting and forecast modeling, quantum computing could be a game changer. The data modeling capabilities of quantum computers are expected to be superior in finding models, performing classifications, and making predictions that are not possible today with conventional computers due to the challenges of complex data structures. Another use case in the world of finance is risk analysis. Risk analysis calculations are difficult because it is difficult to analyze many scenarios. Compliance costs are expected to more than double in the coming years. Financial services institutions are under increasing pressure to balance risk, hedge positions more effectively and perform a wider range of stress tests to comply with regulatory requirements. Today, Monte Carlo simulations — the preferred technique for analyzing the impact of risk and uncertainty in financial models — are limited by the scaling of the estimation error. Quantum computers have the potential to sample data differently by testing more results with greater accuracy, providing quadratic acceleration for these types of simulations.
You can find use cases and tutorials in https://qiskit.org/documentation/finance/ and test quantum algorithms:
· Quantum Amplitude Estimation
· Pricing European Call Options
· Pricing European Put Options
· Pricing Asian Barrier Spreads
· Loading and Processing Stock-Market Time-Series Data
Nature
Molecular modeling allows for discoveries such as more efficient lithium batteries. Quantum computing will empower model atomic interaction much more precisely and at much larger scales. New materials will be able to be used everywhere, whether in consumer products, cars, batteries, etc. Quantum computing will allow molecular orbit calculations to be performed without approximations. A better understanding of the interactions between atoms and molecules will make it possible to discover new drugs. Detailed analysis of DNA sequences will help detect cancer earlier by developing models that will determine how diseases develop. The advantage of quantum will be to analyze in detail on a scale never reached the behavior of molecules. Chemical simulations will allow the discovery of new drugs or better predict protein structures, scenario simulations will better predict the risks of a disease or its spread, the resolution of optimization problems will optimize the chains of distribution of drugs, and finally the use of AI will speed up diagnoses, analyze genetic data more precisely.
You can find use cases and tutorials in https://qiskit.org/documentation/nature/ and test quantum algorithms:
· Sampling the potential energy surface
· Calculating Thermodynamics Observables with a quantum computer
Optimization
The problem of the commercial traveler can be extended in many fields such as energy, telecommunications, logistics, production chains or resource allocation. For example, in sea freight, there is a great complexity in the management of containers from start to finish: loading, conveying, delivering then unloading in several ports in the world is a multi-parameter problem can be addressed by quantum computing. Another application is the optimization of a country’s electricity network, more predictive environmental modeling and the search for energy sources with lower emissions. Aeronautics will also be a source of use cases. For each landing of an airplane, hundreds of operations are set up: crew change, refueling, cleaning the cabin, baggage delivery, or inspections. Each transaction has sub operations. (The refueling requires a tanker available, a truck driver and two people to fill, in advance it must be sure that the tanker is full). So, in total hundreds of operations and that for only one aircraft landing in limited hours. Now, with hundreds of aircraft landing and sometimes delayed flights, the problem is becoming more and more complex. It is then necessary in real time to recalculate everything for all planes.
You can find use cases and tutorials in https://qiskit.org/documentation/optimization/ and test quantum algorithms:
· Converters for Quadratic Programs
· Max-Cut and Traveling Salesman Problem
· Improving Variational Quantum Optimization using CVaR
· Application Classes for Optimization Problems
· Warm-starting quantum optimization
· Using Classical Optimization Solvers and Models with Qiskit Optimization
Machine Learning
Artificial intelligence (AI) that is changing the way businesses operate in fundamental ways is an area bringing new opportunities for progress but also challenges. The capabilities of AI have greatly increased in their ability to interpret and analyze data. AI is also demanding in terms of computing power because of more and more data to process and the complexity of workflows. Machine learning and quantum computing are two technologies that can potentially allow us to solve complex problems, previously untenable, and help accelerate areas like model training or pattern recognition. The future of computing will certainly be made up of classical computing, biologically inspired and quantum computing.
The intersection between quantum computing and machine learning brings considerable attention in recent years and allowed the development of quantum machine learning algorithms such as Quantum-enhanced Support Vector Machine (QSVM), QSVM multiclass classification, variational quantum classifier or qGANs.
It is clear now that quantum computers have the potential to boost the performance of the machine learning algorithms and may contribute to breakthroughs in different fields such as drug discovery or fraud detection. Data can exhibit structures that are hard to identify which can reduce classification accuracy. The idea is to find better patterns within AI/ML processes by leveraging quantum systems that map data to higher dimensions for training and use.
For AI/ML, the potential use cases for quantum computing are the following:
- Chemicals and Petroleum: drilling locations, seismic imaging
- Distribution and logistics: consumer offer recommender, freight forecasting, irregular behaviors
- Financial Services: finance offer recommender, credit and asset scoring, irregular behaviors (fraud)
- Health Care and Life Science: accelerated diagnosis, genomic analysis, clinical trial enhancements
- Manufacturing: quality control, structural design and fluid dynamics
You can find use cases and tutorials in https://qiskit.org/documentation/machine-learning/ and test quantum algorithms:
· Neural Network Classifier & Regressor
· Quantum Kernel Machine Learning
· qGANs for Loading Random Distributions
· Torch Connector and Hybrid QNNs
· Pegasos Quantum Support Vector Classifier
· Quantum Kernel Training for Machine Learning Applications
Conclusion
Future systems will need to integrate quantum computing capabilities to perform specific calculations. As we have seen, many examples are already available and we can test them directly on quantum hardware.
Sources:
- https://qiskit.org
- https://www.ibm.com/quantum-computing/
- The data center of tomorrow is made up of heterogeneous accelerators, Xavier Vasques : https://arxiv.org/abs/2003.10950
