Improving Predictive Power in Quantum Simulation

Neural networks solve quantum mechanics variationally—without needing pre-computed training data Simulating how electrons behave in molecules and materials is essential for designing drugs, batteries, and catalysts. The standard method—density functional theory—has enabled decades of discoveries, but its iterative self-consistent field approach and repeated matrix diagonalization become computationally demanding for large systems. Luqi Dong and coauthors take a fundamentally different approach. Instead of training neural networks to mimic pre-computed results, they train networks to minimize the energy functional directly—turning machine learning into a variational solver. Their model, DeepDM, predicts the density matrix using equivariant graph neural networks, then maps it to the total energy. The network parameters are optimized through backpropagation to minimize energy, not to match labeled examples. This means no pre-computed datasets required. The challenge: density matrices must satisfy strict physical constraints—Hermiticity, particle number conservation, and idempotency. They handle this through a two-stage architecture. First, a network generates an initial density matrix satisfying these constraints. Second, another network applies an exponential transformation that explores the space of valid density matrices while preserving all constraints mathematically. The results match conventional calculations for both molecules (water, methane) and periodic systems (graphene, diamond). More remarkably, models trained only on primitive unit cells generalize accurately to larger supercells—the graph neural network architecture enables this scaling without retraining. The implication: pre-trained models could provide near-converged initial guesses for subsequent calculations, potentially reducing the computational overhead of large-scale quantum simulations significantly. Paper: https://lnkd.in/e9-8jPBb #ArtificialIntelligence #MachineLearning #DeepLearning #QuantumMechanics #DensityFunctionalTheory #ComputationalChemistry #MaterialsScience #NeuralNetworks #Physics #QuantumComputing #AIforScience #GraphNeuralNetworks #ComputationalPhysics #ElectronicStructure #ScientificComputing

"BLIPs: Bayesian Learned Interatomic Potentials" by Dario Coscia, Pim de Haan, Max Welling "Machine Learning Interatomic Potentials (MLIPs) are becoming a central tool in simulation-based chemistry. However, like most deep learning models, MLIPs struggle to make accurate predictions on out-of-distribution data or when trained in a data-scarce regime, both common scenarios in simulation-based chemistry. Moreover, MLIPs do not provide uncertainty estimates by construction, which are fundamental to guide active learning pipelines and to ensure the accuracy of simulation results compared to quantum calculations. To address this shortcoming, we propose BLIPs: Bayesian Learned Interatomic Potentials. BLIP is a scalable, architecture-agnostic variational Bayesian framework for training or fine-tuning MLIPs, built on an adaptive version of Variational Dropout. BLIP delivers well-calibrated uncertainty estimates and minimal computational overhead for energy and forces prediction at inference time, while integrating seamlessly with (equivariant) message-passing architectures. Empirical results on simulation-based computational chemistry tasks demonstrate improved predictive accuracy with respect to standard MLIPs, and trustworthy uncertainty estimates, especially in data-scarse or heavy out-of-distribution regimes. Moreover, fine-tuning pretrained MLIPs with BLIP yields consistent performance gains and calibrated uncertainties." Paper: https://lnkd.in/dYBUkFbu #machinelearning

I'm excited to share our latest work, Demonstration of robust and efficient quantum property learning with shallow shadows, published in Nature Communications! 🎉 📝 Authors: Hong-Ye Hu, Andi Gu, Swarnadeep Majumder, Hang Ren, Yipei Zhang, Derek S. Wang, Yi-Zhuang You, Zlatko Minev, Susanne F. Yelin, Alireza Seif 🔍 Context: Extracting information efficiently from quantum systems is crucial for advancing quantum information processing. Classical shadow tomography offers a powerful technique, but it struggles with noisy, high-dimensional quantum states and complex observables. 🤔 Key Question: Can we overcome noise limitations and improve sample efficiency in quantum state learning, especially for high-weight and non-local observables, using shallow quantum circuits? 💡 Our Findings: We introduce robust shallow shadows—a protocol designed to mitigate noise using Bayesian inference, enabling highly efficient learning of quantum state properties, even in the presence of noise. Our experiments on a 127-qubit superconducting quantum processor confirm the protocol’s practical use, showing up to 5x reduction in sample complexity compared to traditional methods. ✨ Key Takeaways: 1. Noise-resilience: Accurate predictions across diverse quantum state properties. 2. Sample Efficiency: Substantial reduction in sample complexity for high-weight and non-local observables. 3. Scalability: The protocol is well-suited for near-term quantum devices, even with noise. Paper: https://lnkd.in/dW4NJ23Q

> Sharing resource < Nice one: "Hybrid Quantum-Classical Machine Learning Algorithms for Multi-Output Time-Series Forecasting at Utility Scale" by Mackenson Polche, PhD, Varun Puram, Aditi Lal, Weronika Golletz, Joan Étude Arrow, Vardaan Sahgal, Kumar Ghosh, Giorgio Cortiana, Dr. Corey O'Meara Abstract: Multi-output time-series forecasting in energy systems is challenging because of nonlinear dynamics, multi-scale seasonality, and strong dependencies across correlated series. In this work, we investigate two hybrid quantum-classical frameworks for multi-stream time-series forecasting on a real Smart Meter dataset comprising 103 household electricity consumption time-series, with experiments executed on the ibm_marrakesh superconducting quantum processor. The first model, Kernelized Quantum Reservoir Computing with Repeated Measurement (KQRC-RM), combines coupled quantum reservoirs, ancilla-assisted repeated measurement, and kernelized readouts to model temporal dynamics and cross-stream correlations jointly. For a 3-stream time-series input and output, the KQRC-RM model using 114 qubits achieves an MAE of 0.0811 on MPS simulator (36.92\% improvement over its classical analog) whereas performance degrades to an MAE of 0.1524 on hardware. The second, a Projected Quantum Kernel Gaussian Process (QGP), replaces fidelity-based kernels with projected kernels constructed from local reduced-state statistics. Using a topology-aware 100-qubit QGP model to predict 100 multi-output time-series values, we observe 49\% of time-series outputs achieve high-accuracy predictions (MAE <0.15), with an average MAE of 0.082 for this low-error group. The medium-error regime (MAE 0.15-0.35) has an average MAE of 0.229, while the high-error regime (MAE >0.35) has an average MAE of 0.664. Overall, this reduces the average MAE relative to the classical GP baseline by 62.01\% on MPS simulator and 40.37\% on hardware. Together, these results demonstrate the feasibility of hybrid quantum machine learning for multi-input, multi-output time-series forecasting at the 100+ qubit scale on NISQ devices. Link: https://lnkd.in/edRi84XY

Headline: China’s Oceanlite Supercomputer Marries AI and Quantum Science—37 Million Cores Simulate Molecular Quantum Chemistry Introduction: In a milestone achievement, Chinese researchers have fused artificial intelligence with traditional supercomputing to simulate complex quantum chemistry at molecular scale—without using a quantum computer. Using the Oceanlite supercomputer powered by 37 million processing cores, the Sunway team has achieved a feat previously deemed impossible on classical machines. Key Insights: 1. Bridging AI and Quantum Physics Quantum chemistry models the probabilistic behavior of particles like electrons within molecules, governed by the wavefunction (Ψ). Such simulations are normally restricted to small molecules due to the exponential complexity of quantum states. To overcome this barrier, the Sunway team used neural-network quantum states (NNQS), allowing machine learning to approximate molecular wavefunctions with quantum-level accuracy. 2. Record-Breaking Simulation Researchers modeled a molecular system containing 120 spin orbitals—the largest AI-driven quantum chemistry simulation ever conducted on a classical supercomputer. The NNQS trained to predict electron energy distributions and refined itself iteratively until it mirrored true molecular quantum behavior. This approach demonstrates that deep learning frameworks can replicate quantum effects at unprecedented scale. 3. Oceanlite’s Engineering Triumph The experiment ran on the Sunway SW26010-Pro CPU, each chip featuring 384 cores optimized for high-performance computing (HPC). Engineers built a hierarchical communication model where management cores coordinated millions of lightweight compute processing elements (CPEs). Achieved 92% strong scaling and 98% weak scaling efficiency, indicating near-perfect hardware-software synchronization—an exceptional accomplishment in exascale computing. 4. Strategic and Scientific Impact Marks a leap forward for China’s AI and quantum research sectors, blending HPC power with neural architectures. The achievement positions China at the frontier of simulating quantum systems without quantum hardware. Why It Matters: This breakthrough redefines the boundary between classical and quantum computing, offering a path to simulate and design complex molecules—essential for materials science, drug discovery, and clean energy research—using today’s infrastructure. It also signals China’s deepening command of exascale computing and its integration with AI, setting a new global benchmark in scientific computing innovation. I share daily insights with 28,000+ followers and 10,000+ professional contacts across defense, tech, and policy. If this topic resonates, I invite you to connect and continue the conversation. Keith King https://lnkd.in/gHPvUttw

⚛️ Scalable Quantum Reservoir Computing over Distributed Quantum Architectures 📜 Reservoir computing provides an alternative to recurrent neural networks by overcoming the common problems of backpropagation through time and by training only a simple readout layer. The emerging field of quantum computing offers a new computing paradigm that promises to enhance learning through richer feature representations. In this work, we investigate quantum reservoir computing for time-series forecasting. We explore and benchmark four different architectures that combine single or multiple (distributed) reservoirs with single or multiple (distributed) ridge-regression readout layers. We evaluate these architectures using ideal and hardware-informed noisy simulations, and include both hybrid and fully quantum variants, with classical reservoir counterparts serving as a baseline. The results indicate that quantum-enhanced configurations consistently improve forecasting accuracy by reducing the mean absolute error (MAE) and the root mean squared error (RMSE) up to 78.8% and 72.3%, respectively, while distributed architectures effectively enable scaling by utilizing multiple quantum resources in a hardware-agnostic manner. These findings support distributed quantum reservoir computing as a promising, modular approach for forecasting on the quantum platforms of the noisy intermediate-scale quantum (NISQ) era. ℹ️ Liliopoulos et al - 2026

Utilization of SU(2) Symmetry for Efficient Simulation of Quantum Systems Abstract This work investigates variational compilation methods for simulating quantum systems with internal SU(2) symmetry. The central component of the research is the application of the Dynamic Mode Decomposition (DMD) method to extrapolate trained variational circuit parameters beyond the initial optimization range. An approach is proposed for predicting variationally compiled quantum states with a larger number of Trotter steps using extrapolated parameters, eliminating the need for retraining. The efficiency of the method is validated by comparing it with classical Trotterization and the results of variational training. The proposed method demonstrates an effective integration of symmetry-consistent quantum circuit architecture with spectral prediction techniques. The methodology shows promise for scalable modeling of strongly correlated systems, particularly in condensed matter physics problems, such as the Heisenberg model on Kagome lattices

Simulating complex quantum materials has always been one of the biggest challenges in physics. The classical computational cost explodes exponentially with system size and evolution time. A new research paper on arXiv shows that contemporary digital quantum processors are ready to tackle this bottleneck today. Researchers at Q-CTRL successfully executed large-scale digital quantum simulations of the 1D Fermi-Hubbard model using up to 120 qubits on the ibm_boston processor. IBM Quantum Here are the key takeaways from the paper: * Unprecedented Scale: The team simulated up to 60 lattice sites with 90 Trotter steps and over 13,800 two-qubit gates, far beyond exact classical statevector limits. * Real Physical Insights: By tracking defect propagation in a Néel state, they directly observed spin-charge separation, matching theoretical predictions from the Bethe ansatz. * Massive Speedup: At the limits of quantum-classical agreement, the quantum processor ran over 500x to 3000x faster than leading classical tensor-network (TDVP) solvers. * Smart Compilation: They combined pair-interleaved qubit mapping, fSWAP networks, and overhead-free error suppression to achieve deep, high-fidelity circuits. This study proves that pre-fault-tolerant quantum hardware can already deliver fast, accurate, and competitive results for condensed matter physics. What are your thoughts on using near-term quantum processors for material science? Let us know in the comments. #QuantumComputing #Physics #DeepTech #FermiHubbard #QuantumSimulation #TechInnovation

🧬 Quantum Supremacy? Google’s New Quantum Algorithm Could Transform Molecular Simulation Google recently published a Nature paper describing a new quantum algorithm called Quantum Echoes — a technique purpose-built for quantum computers that may become the foundation for faster, more accurate molecular simulations. In simple terms, this algorithm lets researchers simulate how a small change in one part of a molecule (like a protein, RNA, or DNA segment) affects distant regions of the same molecule — something that’s computationally brutal on classical machines. The breakthrough is in scaling: - 🧮 Traditional simulations slow down exponentially as the number of atoms increases. - ⚛️ Quantum Echoes, running on Google’s new Willow quantum chip, scales linearly — achieving roughly a 13,000× speed-up compared to the world’s fastest classical supercomputer. What makes this especially exciting is that few quantum algorithms have ever shown verified speed-ups over classical methods — and most well-known examples (like Shor’s algorithm, which breaks encryption, or Grover’s algorithm, used for database searches) are primarily computer science milestones. Quantum Echoes is one of the first algorithms with direct scientific relevance — with potential applications in biology, chemistry, and drug discovery. It could help refine molecular structures by filling in gaps where experimental techniques struggle: - 🧱 X-ray crystallography gives high-resolution geometry but only in crystal form. - ❄️ Cryo-EM captures large complexes, but flexible regions often go blurry. - 🔬 NMR detects local interactions, but long-range couplings are weak and hard to measure. By simulating the missing connections between distant parts of a molecule, this algorithm could provide extra structural constraints in ambiguous regions, improving model accuracy and bridging the gap between experimental biology and computational modelling. The future of molecular simulation might just be quantum-powered. ⚛️ 🔗 Links to the paper and Google’s announcement are in the comments. #QuantumComputing #MolecularSimulation #Bioinformatics #StructuralBiology #ComputationalBiology #DrugDiscovery #QuantumBiology #NMR #CryoEM #GoogleQuantumAI #WillowChip #NaturePaper #ScienceInnovation #FutureOfBiotech #QuantumAdvantage