Testing

nn

As a funder, EPSRC is interested in understanding how we could best support our research communities?

EPSRC will be generally interested to hear the discussions, and understand what the main research challenges are or will be, and where there may be key infrastructure and/or skills needs for the UK in the coming years. This doesn’t necessarily need to be a separate question, as the challenges and needs may well come out in discussions anyway.

What kind of applications are practical for large-scale quantum/classical hybrid systems?

The concept of quantum/classical hybrid systems is widely accepted because it is believed that in NISQ era classical computation devices need to assist quantum computing devices that can perform only small calculation. However, the idea of quantum/classical hybrid computation incubate a variety of algorithms such as VQE. So, if we can utilize the large scale classical HPC resources with more accurate quantum computers, what kind of practical applications can be realized in the future?

– Scalable pre-/post-processing by HPC for quantum computation.
– Orchestration of noisy quantum devices and noise-free quantum circuit simulation by HPC.
– Quantum/AI workflow processing.

In the near term, how can the research community effectively accelerate the discovery and practical implementation of quantum algorithms to put us on the pathway to achieving quantum advantage? Which applications of QC should be prioritised, and which communities need to collaborate to make this happen?

We believe that quantum advantage is firmly on the horizon but that it will require a broad collaborative effort in to accelerate algorithms development across the research ecosystem. We would be interested to hear the following explored
• Challenges for practical implementations of quantum algorithms in the near-term
• Most promising directions for practical quantum advantage in the next few years
• Specific communities that will need to collaborate to enable advantage.

How to select the most suitable encoding for QML algorithms such as QSVM? Is it possible to identify that by analyzing the data or any other parameters such as hardware parameters.

We are currently exploring Quantum Support Vector Machines (QSVM) and Quantum Neural Networks (QNN), applying them to datasets relevant to various industries. At a smaller scale—particularly with imbalanced datasets from sectors such as banking and insurance—we are observing some promising results. We are now looking to scale these experiments further and would like to understand the future potential of these approaches, the challenges associated with scaling, and their practical applicability in real-world industrial settings.

7958: For fraud detection, can hybrid quantum-classical models deliver a statistically significant uplift over state-of-the-art classical pipelines on today’s quantum hardware?

7930: For Kernel based QML algorithms, presently we compute the Kernel matrix in quantum hardware. To get the Kernel matrix we have to execute O(n^2) circuits, where n is the number of data points. So the number of circuits to be executed becomes a bottleneck for large data sets. What can be a solution to this problem so that we can train a QML model based on a dataset of practical industry scale.

7958: Financial fraud is increasingly complex and we need more effective methods for detection. Quantum kernel methods and quantum neural networks have shown promise in early benchmarks, but it remains unclear whether they can consistently outperform classical ML for these tasks, especially under realistic constraints. Can quantum kernels or quantum graph networks actually beat classical baselines?

7930: We are currently exploring Quantum Support Vector Machines (QSVM) and Quantum Neural Networks (QNN), applying them to datasets relevant to various industries. At a smaller scale—particularly with imbalanced datasets from sectors such as banking and insurance—we are observing some promising results. We are now looking to scale these experiments further and would like to understand the future potential of these approaches, the challenges associated with scaling, and their practical applicability in real-world industrial settings.

7962: What are the prospects of quantum advantage in machine learning before and after fault tolerance?

7924: Which classes of problems (e.g., classification, regression, clustering) are most likely to benefit from Quantum Machine Learning (QML), and which QML algorithms (such as Quantum Neural Networks, Quantum Support Vector Machines etc) are expected to outperform their classical counterparts? What are the key technical and practical challenges that must be addressed to unlock this potential?

7962: A fault-tolerant quantum computer could provide a quantum advantage by learning from (copies of) quantum data. Before fault tolerance, sampling hardness was mentioned as a potential pre-requisite, as well as non-classical simulability of the quantum data encoding or quantum model. Other conjectures refer to the size of the reproducing kernel hilbert space, its spectra (alignment and anti-concentration). More recent developments hint at the role of non-stabilizerness.

In this complex picture, which concept seems more relevant to seeking an advantage on benchmark (e.g. Kaggle) datasets?

7924: We are currently exploring Quantum Support Vector Machines (QSVM) and Quantum Neural Networks (QNN), applying them to datasets relevant to various industries. At a smaller scale—particularly with imbalanced datasets from sectors such as banking and insurance—we are observing some promising results. We are now looking to scale these experiments further and would like to understand the future potential of these approaches, the challenges associated with scaling, and their practical applicability in real-world industrial settings.

How can research and industry inform policymakers about quantum computing before making important decisions to ensure that it is used for responsible applications?

Quantum computing is a technology that has the potential to have both positive and negative impacts on society. Policy frameworks for similarly disruptive technologies in the past have struggled to keep up with the rate at which these technologies are developed, meaning that in some cases, little can be done to prevent misuse. To ensure this is not the case for quantum computing, it is important that policy frameworks are developed in advance, meaning that policymakers need to be well informed about the technology. However, quantum computing can at times be difficult to understand and often suffers from misinformed or misleading media outlets. What can be done to make sure that policymakers have the right knowledge of quantum computing to inform their decision-making?

What can be done to prevent quantum computing from furthering the digital divide between the developed world and underserved geographies?

Large private sector organisations have already invested heavily in quantum computing technology, with certain geographies investing more than others. To ensure that quantum computing does not widen the digital divide, resulting in higher levels of global inequality, it is important to ensure that people from underserved geographies are educated and have access to quantum computing technologies. What can be done to make sure that this is the case?

How will compilers need to adapt to incorporate new fault tolerant components?

The next generation of devices will not have all of the components necessary for a fully fault-tolerant quantum computer. However, they will have some components. What needs to be prioritised to enable algorithms researchers to explore algorithms beyond the current variational models?

How can we address training requirements for regulators and financial services firms to support long-term adoption of QC?

Upskilling the quantum ecosystem is a core component of the government missions. Regulators and financial services firms need to be taught more about quantum technologies and quantum researchers that want to find applications need to know the business models and regulatory frameworks they are addressing. What are some key, innovative, collective actions that could be taken to ensure the UK delivers on this upskilling component of its missions?

Can can Quantum Computing and Generative AI solve some of the critical challenges of modern industry?

The discussion highlights QTNM’s unique approach of merging generative AI with cutting-edge quantum solutions and hardware to address some of the world’s most complex challenges – from accelerating the discovery of new drugs and targeted delivery mechanisms to improving carbon capture, and efficiency of hydrogen fuel cells and batteries active materials. In close collaboration with subject matter experts from world-leading institutions and industrial partners – including BMW, Airbus, Amgen, QSL, the Forgan Group (University of Glasgow), TotalEnergies, and others – QTNM is driving pioneering work that has already led to published results and continues to advance toward first commercial value. The purpose of discussion to identify the challenges of new materials discovery and the timelines for practical advantage, that Generative AI and Quantum could deliver together. References:
https://www.nature.com/articles/s41524-024-01460-x
https://epjquantumtechnology.springeropen.com/articles/10.1140/epjqt/s40507-022-00155-w

What Quantum Comptutational techniques show most promise in the fight against pathogens specificallyin studying the genetics of bacteria ( anti-microbrial resistance) and viruses ( the development of vaccines)

Follow up study to QBA/QSL Proof of concept summarised here: https://gtr.ukri.org/projects?ref=10083188

How can quantum-accelerated optimisation be embedded into existing CAE / digital-twin workflows to cut design-cycle time and what benchmark problems would convincingly demonstrate a quantum advantage over leading classical solvers?

Engineering optimization problems—like turbine blade design or smart-factory scheduling—often require high-dimensional, multi-physics simulations that strain classical resources. Embedding quantum algorithms into digital-twin or CAE (Computer-Aided Engineering) pipelines could reduce design cycles, improve energy efficiency, or uncover novel solutions. This question aims to identify which industrial workflows are most likely to benefit from quantum acceleration in the next 3–5 years.

7958: For fraud detection, can hybrid quantum-classical models deliver a statistically significant uplift over state-of-the-art classical pipelines on today’s quantum hardware?

7930: For Kernel based QML algorithms, presently we compute the Kernel matrix in quantum hardware. To get the Kernel matrix we have to execute O(n^2) circuits, where n is the number of data points. So the number of circuits to be executed becomes a bottleneck for large data sets. What can be a solution to this problem so that we can train a QML model based on a dataset of practical industry scale.

7958: Financial fraud is increasingly complex and we need more effective methods for detection. Quantum kernel methods and quantum neural networks have shown promise in early benchmarks, but it remains unclear whether they can consistently outperform classical ML for these tasks, especially under realistic constraints. Can quantum kernels or quantum graph networks actually beat classical baselines?

7930: We are currently exploring Quantum Support Vector Machines (QSVM) and Quantum Neural Networks (QNN), applying them to datasets relevant to various industries. At a smaller scale—particularly with imbalanced datasets from sectors such as banking and insurance—we are observing some promising results. We are now looking to scale these experiments further and would like to understand the future potential of these approaches, the challenges associated with scaling, and their practical applicability in real-world industrial settings.

7962: What are the prospects of quantum advantage in machine learning before and after fault tolerance?

7924: Which classes of problems (e.g., classification, regression, clustering) are most likely to benefit from Quantum Machine Learning (QML), and which QML algorithms (such as Quantum Neural Networks, Quantum Support Vector Machines etc) are expected to outperform their classical counterparts? What are the key technical and practical challenges that must be addressed to unlock this potential?

7962: A fault-tolerant quantum computer could provide a quantum advantage by learning from (copies of) quantum data. Before fault tolerance, sampling hardness was mentioned as a potential pre-requisite, as well as non-classical simulability of the quantum data encoding or quantum model. Other conjectures refer to the size of the reproducing kernel hilbert space, its spectra (alignment and anti-concentration). More recent developments hint at the role of non-stabilizerness.

In this complex picture, which concept seems more relevant to seeking an advantage on benchmark (e.g. Kaggle) datasets?

7924: We are currently exploring Quantum Support Vector Machines (QSVM) and Quantum Neural Networks (QNN), applying them to datasets relevant to various industries. At a smaller scale—particularly with imbalanced datasets from sectors such as banking and insurance—we are observing some promising results. We are now looking to scale these experiments further and would like to understand the future potential of these approaches, the challenges associated with scaling, and their practical applicability in real-world industrial settings.

What can we usefully do with a MegaQuOP?

I can rephrase this question as, what is the largest classical state we can load and then measure on a MegaQuOP device? I posit that this is likely to be of the order of 5-6 qubits. I also posit that this is also a useful thing to do.

What is the current status of Quantum Error Correction and Database search quantum algorithms that can be used for analysis of biological samples?

Given the amount of data required to analyse and the usage of computational resources, Quantum Computing can represent a breakthrough in analysis of data pipelines from experimental data to characterise proteins and peptides from biological samples. Currently, deep learning methods and database search algorithms are used as part of the data pipeline. But the library of peptides is extremely large. So, given the number of qubits required and the process for identification, Quantum Error Correction could have a strong impact. We want to check the viability of using quantum computing when there is a large biological database to compare experimental data.

What are some of the main challenges that quantum circuits and quantum algorithms have when applied to modelling and simulation scenarios?

We are interested in looking at the viability of using quantum computing, quantum simulation on a multiplicity of scenarios in which modelling of complex systems in different fields such as proteomics, environmental science, agriculture. We work with statistical modelling and modelling and simulation techniques, but some of them are computationally expensive and also have lots of variables. We would like to discuss how feasible is to use quantum computing so that we can optimise our models.