Michele Grossi

Hi, the aqora team is working on this.

Thank you for the comment. Yes, I agree that R^2 seems to work better.

Yes, I think this should work.

Hi there, apologise for this inconvenience. We do understand this, the problem is the higher parton density functions that can cause 'd_sigma' to become negative. We propose the following: take the absolute value of this to calculate the MAPE score.

This is a fair question. However we need to take into account we have multiple technologies for building a quantum computer. Superconducting is of course the one that requires extremely low temperature. There are many studies about energy footprint of QC, but the computation power of qubit is scaling better than classic ones, especially if you consider big data centres. Though cooling requires low temperatures (~millikelvin range), quantum computers may still be more energy-efficient than classical supercomputers for specific tasks, reducing overall computational carbon footprints in the long run. Overall, their ability to solve complex sustainability challenges (e.g. better catalysts for carbon capture, developing low-energy fertilizers, discovery of new materials) could ultimately offset these costs, making them a valuable tool in combating climate change.

thank you for the question. At CERN we are running CERN QUANTUM TECHNOLOGY INITIATIVE, where quantum computing represents 1 competence center. See here for additional details and the other Cocs https://quantum.cern/research In general, our mission is to integrate with the EU and US HPC+QC infrastructures and take part in the design and deployment of hybrid computing solutions in support of several scientific use cases. We have PhD students and Post Docs that work on different aspects of computing, from theoretical foundations of QML, to more application oriented algorithms. We also host conferences, this year we had QT4HEP2025 where you can find many of the topics we work on: https://indico.cern.ch/event/1433194/overview

Thank you for the question. Quantum technologies are evolving so fast, this affects the way we think about algorithms. In particular, there have been key results in hardware improvements as well as in quantum error correction. See for example the Google's Willow Processor (2024) where they achieved below-threshold error correction, improving qubit stability. IBM is actively working on demonstrate the first quantum-centric supercomputer by integrating modular processors, middleware, and quantum communication. They also enhance the quality, execution, speed, and parallelization of quantum circuits.

In hybrid quantum-classical methods there is continuous effort in reducing resource demands, improving optimization routines and include tensor network alternatives.

In HEP we have collected most interesting results here: REF .

Improvements in the hardware quality allow to run and test more complex circuits in terms of circuit depth, that means getting into more realistic examples.

We were able to perform a real calculation of a non trivial cross section on quantum hardware 10.1088/2058-9565/ada9c5. Regarding generative model, we have interesting direction for QML looking into QBM: https://arxiv.org/abs/2410.16363

This is true for quantum field theory calculation (mainly lattice based) but also for data analysis and data generation. We can get to better quantum field theory simulations via enhanced error mitigation

See https://www.nature.com/articles/s41567-024-02723-6 and https://arxiv.org/abs/2411.13161