AI Model Development for Quantum Systems

🔗✨ Exploring the Future of Quantum Computing with Physics-Informed Neural Networks (PINNs) ✨🔗 Excited to highlight the pioneering work by Stefano Markidis that dives deep into the potential of Quantum Physics-Informed Neural Networks (Quantum PINNs) for solving differential equations on hybrid CPU-QPU systems! 📘 What’s this about? Physics-Informed Neural Networks (PINNs) have proven their versatility in addressing scientific computing challenges. This study extends PINNs into the quantum realm using Continuous Variable (CV) Quantum Computing, offering a new approach to solving Partial Differential Equations (PDEs) with quantum hardware. Key Highlights: ✅ Quantum Meets Physics: The framework combines CV quantum neural networks with classical methods to tackle PDEs like the 1D Poisson equation. ✅ Optimizer Insights: Traditional optimizers like SGD outperformed adaptive methods in this quantum landscape, highlighting the unique challenges of quantum optimization. ✅ Scalability: Explores batch processing and neural network depth for more effective performance on quantum systems. ✅ Programming Ease: Tools like Strawberry Fields and TensorFlow simplify the integration of quantum and classical computations. 💡 Why it matters: This research doesn't just apply PINNs to quantum computing—it highlights the differences between classical and quantum approaches, paving the way for advancements in quantum PINN solvers and their real-world applications in computational physics, electromagnetics, and more. 📖 Dive deeper: Access the full study here: https://lnkd.in/dZm3F3CR Source code available: https://lnkd.in/dAsXxnbN What are your thoughts on combining quantum computing with AI for scientific breakthroughs? Let’s discuss! 🚀 #QuantumComputing #PhysicsInformedNeuralNetworks #ScientificComputing #HybridAI #PDEsolvers #Innovation

Training ML on ground-state physics using quantum computing experiments has so far been limited to small systems or highly structured states. In our new preprint, we push this paradigm to 2D systems of up to 115 qubits. Out on the arXiv today https://lnkd.in/eUzgeWHH, we demonstrate learning on ground states of the interacting 2D Heisenberg XXZ model. For system sizes of 57 and 115 qubits, we generate a dataset of 1-, 2- and 12-local Pauli observables that agree with DMRG to within a few percent across most of the antiferromagnetic phase. Training neural networks on this data, we find they can accurately predict specific spatial patterns of observables across the lattice for previously unseen Hamiltonians. The models generalise the behaviour of the underlying quantum experiments, both within the training distribution and in an out-of-distribution regime approaching the phase boundary. Ultimately, our results help us push towards a regime where AI may be trained on data from quantum processors that are fundamentally beyond the reach of classical approximation methods. A huge thanks to my co-authors across IBM Quantum, the University of Oxford, and the STFC Hartree Centre including Freya Shah, Minjun Jeon, M. Emre Şahin, Christa Zoufal, Kunal Sharma IBM STFC

🚀 New Paper on arXiv! I’m excited to share our latest work: “Learning to Program Quantum Measurements for Machine Learning” 📌 arXiv: https://lnkd.in/euRhBQJM 👥 With Huan-Hsin Tseng (Brookhaven National Lab), Hsin-Yi Lin (Seton Hall University), and Shinjae Yoo (BNL) In this paper, we challenge a long-standing limitation in quantum machine learning: static measurements. Most QML models rely on fixed observables (e.g., Pauli-Z), limiting the expressivity of the output space. We take this one step further--by making the quantum observable (Hermitian matrix) a learnable, input-conditioned component, programmed dynamically by a neural network. 🧠 Our approach integrates: 1. A Fast Weight Programmer (FWP) that generates both VQC rotation parameters and quantum observables 2. A differentiable, end-to-end architecture for measurement programming 3. A geometric formulation based on Hermitian fiber bundles to describe quantum measurements over data manifolds 🧪 Experiments on noisy datasets (make_moons, make_circles, and high-dimensional classification) show that our dual-generator model outperforms all traditional baselines—achieving faster convergence, higher accuracy, and stronger generalization even under severe noise. We believe this work opens the door to adaptive quantum measurements and paves the way toward more expressive and robust QML models. If you're working on QML, differentiable quantum programming, or quantum meta-learning, I’d love to connect! #QuantumMachineLearning #QuantumComputing #QML #FastWeightProgrammer #DifferentiableQuantumProgramming #arXiv #HybridAI #AI #Quantum

Quantum computing promises to making LLMs more efficient. And it's already working on real hardware. Efficient fine-tuning of large language models remains a critical bottleneck in AI development, with most researchers focused on purely classical computing approaches. A new paper from Chinese researchers demonstrates how quantum computing principles can dramatically reduce the parameters needed while improving model performance. The team introduces Quantum Weighted Tensor Hybrid Network (QWTHN), which combines quantum neural networks with tensor decomposition techniques to overcome the expressive limitations of traditional Low-Rank Adaptation (LoRA). By leveraging quantum state superposition and entanglement, their approach achieves remarkable efficiency: reducing trainable parameters by 76% while simultaneously improving performance by up to 15% on benchmark datasets. Most importantly, this isn't just theoretical - they've successfully implemented inference on actual quantum computing hardware. This represents a tangible advancement in making quantum computing practical for AI applications, demonstrating that even current-generation quantum devices can enhance the capabilities of billion-parameter language models. The integration of quantum techniques into traditional deep learning frameworks might become standard practice for resource-efficient AI development in the future. More on Quantum Hybrid Networks and other AI highlights in this week's LLM Watch:

A new preprint, “Learning ground state observables from quantum computing experiments,” shows how machine learning can predict properties of interacting many-body systems from data generated on quantum processors, on systems of up to 115 qubits: https://lnkd.in/euAitfbd In collaboration with University of Oxford and STFC Hartree Centre, researchers from IBM developed a basis-optimization technique that combines sample-based quantum diagonalization (SQD) with observable backpropagation, allowing the low-energy subspace to be represented using entangled basis states. Using this workflow, the researchers could study the two-dimensional Heisenberg XXZ model and generate experimental training data across the antiferromagnetic phase. The dataset includes local observables, two-point correlations, and 12-body loop observables, with the same framework extending in principle to global observables. Classical neural networks trained on these data accurately predict spatially resolved observables at Hamiltonian parameters not included during training, including beyond the training range. This work demonstrates an end-to-end workflow combining quantum experiments, high-performance computing, and classical machine learning. More broadly, it shows how quantum computing and AI can complement one another, with quantum processors producing data about complex physical systems and AI using that information to accelerate scientific discovery.

***new preprint on quantum GenAI** How do you learn a complex probability distribution from examples, and then generate new samples from it? This becomes difficult when variables are continuous, correlated, multi-modal, or have sharp features that are easy for generative models to smooth away. Together with collaborators at CERN, we have developed an algorithmic framework for learning and sampling such distributions with quantum circuits: https://lnkd.in/e3QPaXv2 Huge thanks to Cenk Tüysüz and Michele Grossi for pushing this paper forward. It has been a wonderful effort. The idea is a quantum Fourier generative model with a bipartite structure that allows an efficient Monte Carlo estimation of a likelihood-based loss. The model can be trained classically at the scale of 1,000 qubits on a single GPU, with evaluations taking below a second. The workflow is then simple: train on classical hardware, deploy on quantum hardware for sampling. We test the approach on distributions with multiple peaks, oscillations and heavy tails; on correlated financial data; and on multivariate distributions relevant to high-energy physics. Compared with IQP-based generative models trained using MMD loss, the likelihood approach captures much more of the relevant high-order structure and correlations. On difficult multi-modal benchmarks, it also preserves separated modes rather than blurring them together, performing competitively with tuned normalizing flows and diffusion baselines. The cherry on top was the hardware results. We deployed trained circuits on IBM Quantum’s superconducting processor and sampled distributions including multi-modal targets and a truncated Lévy distribution. Even without error mitigation, the structure is well captured. Certainly many challenges remain. While training is no longer the fundamental bottleneck, reliable sampling from large trained circuits is, as well as search for suitable applications. Happy to be an early developer of train-on-classical, deploy-on-quantum approach, and excited to see what we can do next. Big thanks to EPSRC and QCi3 Hub for supporting me in this. Preprint: Quantum Fourier Generative Models Trainable at Large Scale arXiv:2606.28483 & SciRate https://lnkd.in/e4nbT-GB #QuantumComputing #QuantumMachineLearning #GenerativeAI #CERN

I'm really happy with the rapid development of CUDA-Q QEC, our toolkit for quantum error correction. QEC is an incredibly rich and fast-moving field, and in CUDA-Q QEC we aim to provide a platform with a diverse set of accelerated decoders, AI infrastructure, tools to enable researchers to develop and test their own codes, decoders, and architectures, hopefully even better than our own! As we dig deeper into the problem of scalable QEC, the benefits of GPUs and AI have become much clearer. We started with research tools, for simulation and offline decoding, which is still an important capability. Now with the 0.5.0 release we also provide the infrastructure for real-time decoding, where syndrome processing occurs concurrently with quantum operations. This release also introduces GPU-accelerated algorithmic decoders like RelayBP, a promising approach developed in the past year that aims to overcome the convergence limitations of traditional belief propagation. For scenarios demanding maximum throughput, we have integrated a TensorRT-based inference engine that allows researchers to deploy custom AI decoders trained in frameworks like PyTorch and exported to ONNX directly into the quantum control loop. To address the complexities of continuous system operation, we added sliding window decoders that handle circuit-level noise across multiple rounds without assuming temporal periodicity. These tools are designed to be hardware-agnostic and scalable, supporting our partners across the ecosystem who are building the first generation of reliable logical qubits. Check out the full technical breakdown in our latest developer blog by Kevin Mato, Scott Thornton, Ph.D., Melody Ren, Ben Howe, and Tom L. https://lnkd.in/gvC__zRd

Really happy to see the official publication today of our paper in Nature Machine Intelligence: "Machine Learning for Practical Quantum Error Mitigation" Haoran Liao, Derek S. Wang, Iskandar Sitdikov, Ciro Salcedo, Alireza Seif, Zlatko Minev 🔍 Context: Quantum computers progress to outperform classical supercomputers, but quantum errors remain the primary obstacle. Quantum error mitigation offers a solution but at the high cost of added runtime. 🤔 Key Question: Can classical machine learning help us overcome errors in today's quantum computers by lowering mitigation overheads, in practice, on real hardware, at the 100 qubit+ scale? 🔬 Our Findings: Using both simulations and experiments on state-of-art quantum computers (up to 100 qubits), we find that machine learning for quantum error mitigation (ML-QEM) can: - Significantly reduce overheads. - Maintain or even outperform the accuracy of traditional methods. - Deliver nearly noise-free results for quantum algorithms. We tested multiple machine learning models on various quantum circuits and noise profiles. And, by leveraging ML-QEM, we were able to mimic conventional mitigation results for large quantum circuits, but with much less overhead. 🌟 Conclusion: Our research underscores the potential synergy between classical hashtag#ML and hashtag#AI and quantum computing. We're excited about the prospects and further research! 🙌 Big thanks to the dream team and many folks who contributed! Let’s share and discuss the implications of this exciting work! 🌟👇 📄 Paper: Nature Machine Intelligence https://lnkd.in/dGYzC3fq 🔓 Free access: View the paper here https://lnkd.in/dN222X7D 📚 Preprint on arXiv https://lnkd.in/dGbzjtjA 👩💻 Code Repository: Explore on GitHub https://lnkd.in/dcn-xPtm 🎥 Seminar: Watch hashtag#IBM @Qiskit on YouTube here https://lnkd.in/dEPRcMVK https://lnkd.in/e7JFgc3J

“Before you can use a quantum computer, you first need to be able to turn it on.” Research that I carried out during my PhD at Oxford has brought us closer to that goal. I'm pleased to share that our paper titled “Cross-architecture tuning of silicon and SiGe-based quantum devices using machine learning” has been published in Nature Scientific Reports. We developed CATSAI (pronounced: Cats-eye), an algorithm capable of tuning three different semiconductor quantum devices—silicon finFET, Ge/Si nanowire, and Ge/SiGe heterostructure— to double quantum dots, using a single approach Forming double quantum dots in these devices is a key step towards creating qubits, the essential building blocks of quantum computers. Not long ago, it was thought that each device type would need its own specialized algorithm. CATSAI changes that by tuning different devices and revealing the complex hypersurfaces that separate regions where current flows from those where it’s blocked. In some cases, finding a double quantum dot is like finding a needle in a haystack—sometimes in just 0.002% of the search space. CATSAI does this on the order of minutes —far quicker than what would typically be possible manually. I remember when I first tried to tune a double quantum dot at the start of my PhD - it took me two weeks. That became the last time I tried to do it by hand. CATSAI relies on two key strategies: 1. Training a machine learning model to recognize single quantum dot features. 2. Leveraging reliable data on where these single dots are located in voltage space to narrow down the search for double quantum dots. This work wouldn’t have been possible without the support of our co-authors and collaborators at IST Austria and the University of Basel. Special thanks to Natalia Ares, who supervised my PhD research and provided invaluable guidance and support throughout this project. I’m also grateful for the opportunity she gave me to work with such an amazing team and technology. Interested in learning more? You can read the full paper here: https://lnkd.in/e7Vz8We9 The possibilities ahead are vast, and I’m eager to see where AI software for semiconductor quantum devices takes us next!