Quantum Research Classification Methods

I’d like to draw your attention to a new paper on arXiv, “Shallow-circuit Supervised Learning on a Quantum Processor”, from IBM and Qognitive that develops a Hamiltonian-based framework for quantum machine learning. Instead of fixed amplitude or angle encodings used in many prior approaches, our method learns a local Hamiltonian embedding for classical data. https://lnkd.in/ejcxYstW We are very interested in new approaches to QML as we deal with recurring bottlenecks like expensive classical data loading and difficult training dynamics in parameterized circuit models. Here, both the feature operators and the label operator are learned during training, with predictions obtained from measurements on an approximate ground state. This aims to avoid those bottlenecks. A key enabler is Sample-based Krylov Quantum Diagonalization (SKQD), which approximates low-energy states by sampling from time-evolved Krylov states and then diagonalizing the Hamiltonian in the sampled subspace. SKQD was recently employed to estimate low-energy properties of impurity models (https://lnkd.in/epwCrG5R). In our setting, restricting to 2-local Hamiltonian embeddings keeps the required time-evolution circuits relatively shallow, which helps make the approach practical on current quantum processors. The team demonstrates end-to-end training on IBM Heron processor up to 50 qubits, with non-vanishing gradients and strong proof-of-concept performance on a binary classification task. There are many exciting next steps here, including testing on broader datasets, using more expressive operator ansatz, and performing systematic comparisons to strong classical baselines to pinpoint when Hamiltonian-based encodings offer the right inductive bias. I encourage the community to try out this approach and explore where it can be extended in meaningful ways.

> Sharing Resource < Interesting benchmark for finance: "Quantum vs. Classical Machine Learning: A Benchmark Study for Financial Prediction" by Rehan Ahmad, Muhammad Kashif, Nouhaila I., Muhammad Shafique Abstract: In this paper, we present a reproducible benchmarking framework that systematically compares QML models with architecture-matched classical counterparts across three financial tasks: (i) directional return prediction on U.S. and Turkish equities, (ii) live-trading simulation with Quantum LSTMs versus classical LSTMs on the S\&P 500, and (iii) realized volatility forecasting using Quantum Support Vector Regression. By standardizing data splits, features, and evaluation metrics, our study provides a fair assessment of when current-generation QML models can match or exceed classical methods. Our results reveal that quantum approaches show performance gains when data structure and circuit design are well aligned. In directional classification, hybrid quantum neural networks surpass the parameter-matched ANN by \textbf{+3.8 AUC} and \textbf{+3.4 accuracy points} on \texttt{AAPL} stock and by \textbf{+4.9 AUC} and \textbf{+3.6 accuracy points} on Turkish stock \texttt{KCHOL}. In live trading, the QLSTM achieves higher risk-adjusted returns in \textbf{two of four} S\&P~500 regimes. For volatility forecasting, an angle-encoded QSVR attains the \textbf{lowest QLIKE} on \texttt{KCHOL} and remains within ∼ 0.02-0.04 QLIKE of the best classical kernels on \texttt{S\&P~500} and \texttt{AAPL}. Our benchmarking framework clearly identifies the scenarios where current QML architectures offer tangible improvements and where established classical methods continue to dominate. Link: https://lnkd.in/e4WUdr-n #quantummachinelearning #machinelearning #research #paper #benchmark #finance

⚛️ Parallel Data Processing in Quantum Machine Learning 🧾 We propose a Quantum Machine Learning (QML) framework that leverages quantum parallelism to process entire training datasets in a single quantum operation, addressing the computational bottleneck of sequential data processing in both classical and quantum settings. Building on the structural analogy between feature extraction in foundational quantum algorithms and parameter optimization in QML, we embed a standard parameterized quantum circuit into an integrated architecture that encodes all training samples into a quantum superposition and applies classification in parallel. This approach reduces the theoretical complexity of loss function evaluation from O(N^2) in conventional QML training to O(N), where N is the dataset size. Numerical simulations on multiple binary and multi-class classification datasets demonstrate that our method achieves classification accuracy comparable to conventional circuits while offering substantial training time savings. These results highlight the potential of quantum-parallel data processing as a scalable pathway to efficient QML implementations. ℹ️ Ramezani et al - 2025

The study "Benchmarking MedMNIST Dataset on Real Quantum Hardware" presents several important outcomes. This research provides an interesting approach using quantum hardware for inference. Key outcomes of this study: * The QML workflow that the researchers propose first involves classical preprocessing of medical images followed by training device-aware quantum circuits on classical hardware using noiseless simulators. The best-performing trained models are then transpiled and run on IBM quantum hardware with error suppression and mitigation techniques. Finally, the results from the quantum hardware are classically post-processed to obtain the classification labels. * The experimental results on several MedMNIST datasets (PneumoniaMNIST, BreastMNIST, OCTMNIST, RetinaMNIST, DermaMNIST, BloodMNIST, PathMNIST, and OrganSMNIST) establish an interesting benchmark. * The study found that even with a reduced feature set compared to classical methods, the QML models achieved promising classification accuracy and AUC scores on real quantum hardware for various MedMNIST datasets. For instance, the QML model achieved an accuracy of 85.4% and an AUC of 82.2% on the 2-class PneumoniaMNIST dataset after applying error suppression and mitigation. * A comparison with classical machine learning (ML) benchmarks provided in the MedMNIST dataset documentation showed that while the QML results do not consistently outperform the best classical models (which use full-resolution images), they represent a significant step for QML given the hardware limitations and reduced input features. Notably, the QML model's accuracy on PneumoniaMNIST matched that of some classical ResNet models. * A further comparative analysis by training classical ML baselines (ResNet-18 and ResNet-50) using the same reduced 8x8 input features as the QML model on the 5-class RetinaMNIST dataset demonstrated a potential quantum benefit. The QML model outperformed both ResNet architectures in terms of classification accuracy and AUC, despite having significantly fewer trainable parameters, suggesting advantages in both performance and computational complexity in this specific comparison. Overall, this study demonstrates the feasibility and potential of using purely quantum models on current noisy quantum hardware for medical image classification, setting a benchmark and highlighting the impact of device-aware circuit design and error mitigation techniques. Here the article: https://lnkd.in/dsrCGaqq Here the GitHub repo: https://lnkd.in/dqYTwXqF #quantum #qml #datascience #machinelearning

Exciting work from Caltech, Google Quantum AI, MIT, and Oratomic on quantum advantage for classical machine learning. The long standing question: can quantum computers offer a rigorous advantage in large scale classical data processing, not just specialized problems like cryptography or quantum simulation? This paper gives rigorous results for formalized machine learning tasks. In the benchmarks they report, a quantum computer with fewer than 60 logical qubits performs classification and dimension reduction on massive datasets using 4 to 6 orders of magnitude less memory than the classical and QRAM based baselines in the paper. The key idea is quantum oracle sketching. Instead of loading an entire dataset into quantum memory, it streams classical samples one at a time, applies small quantum rotations, and discards each sample immediately. These operations coherently build an approximate quantum oracle that can then be used in downstream quantum algorithms. The authors present numerical experiments on IMDb sentiment analysis and single cell RNA sequencing that are consistent with the theory. What makes this notable: - A provable quantum memory advantage for classification and dimension reduction - The advantage is framed as a theorem under the paper's learning model, not just a conjecture or empirical trend - The approach is designed to work with streaming, noisy, and time varying classical data Read the paper here: https://lnkd.in/g77PuZzQ

I’m excited to share that our recent paper, "Quantum Machine Learning With Limited Data: A Remote Sensing Perspective," has been published at IGARSS 2025! 🚀 Ever wondered if quantum computers can finally crack the remote sensing data bottleneck?  Honestly, it seems we’re closer than ever.  🛰️⚛️ In Remote Sensing, obtaining extensive ground truth data is often a massive bottleneck. We wanted to see if Quantum Machine Learning (QML) could offer a solution. The Experiment: We compared a Quantum Support Vector Machine (QSVM) against a Classical SVM (CSVM) to classify Holm Oak trees using PRISMA Hyperspectral Imagery. Key Findings: ✅ Less Data, More Accuracy: Our QSVM model outperformed the classical approach by 5% accuracyon limited datasets (just 50 samples). ✅ High Dimensionality: With 12+ qubits, the quantum kernel achieved perfect classification accuracy (1.0) on complex spectral data. ⚠️ The Trade-off: While accurate, QML is currently much slower in training/prediction. The Takeaway: We propose that while QML isn't ready to classify millions of pixels in real-time yet, it has massive potential as a "Pseudo-Labeler." It can learn from tiny datasets to generate labels for larger classical models effectively bridging the gap where ground truth is scarce. A huge thank you to my co-authors Zafer YILMAZ, Saraah Imran, Mohamad Alipour, and Ertugrul Taciroglu. 📄 Read the full paper here: https://lnkd.in/gXsZr8uC #QuantumComputing #MachineLearning #RemoteSensing #IGARSS2025 #HyperspectralImaging #Qiskit #Research #UCLA #OpalAI

Check out this research on the scalable parameterized quantum circuits classifier (SPQCC) by Ding, X., Song, Z., Xu, J. et al., published in Scientific Reports #Nature. This approach addresses the limitations of conventional parameterized quantum circuits (PQC) in multi-category classification tasks, achieving state-of-the-art results. Key Highlights: 🔹 Fast convergence of the classifier 🔹 Parallel execution on identical quantum machines 🔹 Reduced complexity in classifier design 🔹 State-of-the-art simulation results on the MNIST Dataset 🔹 Comparable performance to classical classifiers with fewer parameters Their novel methodology involves per-channel PQC execution, combining measurements as outputs, and optimizing trainable parameters to minimize cross-entropy loss, leading to rapid convergence and enhanced classification accuracy. Explore the open-source research paper and the methodology leading to the results: https://lnkd.in/dT68bR8Q Let’s continue pushing the boundaries of what’s possible with quantum technology! #QuantumComputing #MachineLearning #Innovation #QuantumTech #Research #OpenSource #AI #MNIST Citation: Ding, X., Song, Z., Xu, J. et al. Scalable parameterized quantum circuits classifier. Sci Rep14, 15886 (2024). https://lnkd.in/dT68bR8Q

“A single-qubit quantum circuit which can implement arbitrary unitary operations can be used as a universal classifier much like a single hidden-layer Neural Network.” Ahmed, S., Quantum Computing Researcher. Introduction: A paper titled “Data re-uploading for a universal quantum classifier” by Pérez-Salinas, A., et al. in 2020 detailed how to load a higher dimensional data point represented by x, and breaking it into sets of parameters U(x). Matrix multiplication in circuits represent a linear component, while collapsing the quantum state into 0 or 1 represents the nonlinear component, analogous to classical neural networks. A cost function representing the sum of the fidelities can then be used with an optimizer for determination of optimal weights for classification. Experimental: The author identified the appropriate layers for the model, and then likely adjusted the batch size and learning rate for the fastest run times (Tables 1-3). Adjusting layers did not improve the original method, but increasing the batch size improved accuracy by 2.7% (Table 2); while only decreasing the learning rate increased accuracy by 5.9% (Table 3). Increasing dataset size by 10x yielded similar results, while a 100x increase decreased accuracy, and increased runtime by several hours (Table 4). Triplicate studies were identical for the author’s model, and also for three runs in a revised method. Notebooks are available through GitHub Medical-Quantum-Machine-Learning, Code, PennyLane. Action: The goal of using rich single qubit information to better classify datasets will likely require greater insight into how larger datasets interact with rotational gates. New PQC architectures and revisions to the way quantum parameters are trained likely also need improvements. The Ahmed, S. demo also detailed how more than one qubit can be used to create a Universal Function Approximator. Watch, N., et al. have also successfully used multi-dimensional qudits to “successfully learn highly non-linear decision boundaries of classification problems such as the MNIST digit recognition problem.” Single qubit (binary class) or single qudit (multi-class) methods deserve additional research, as limiting the number of qubits or qudits drastically reduces RAM and Runtime requirements in quantum simulators for larger dataset and classical model incorporations. References: 1) https://lnkd.in/gbjttnjm 2) https://lnkd.in/gZkdvzie 3) https://lnkd.in/gZKG9QiF 4) https://lnkd.in/gyxWXXXr 5) https://lnkd.in/gKwaM75z 6) https://lnkd.in/gsmB8QcZ

I have been exploring how classical deep learning models and quantum circuits can be combined to solve machine learning problems more effectively. Rather than replacing classical approaches, this work focuses on leveraging the strengths of both paradigms. Below is a high-level overview of a hybrid quantum–classical classifier trained on the MNIST dataset (binary classification: digits 0 vs 1). 𝗔𝗿𝗰𝗵𝗶𝘁𝗲𝗰𝘁𝘂𝗿𝗲 𝗢𝘃𝗲𝗿𝘃𝗶𝗲𝘄 𝗖𝗹𝗮𝘀𝘀𝗶𝗰𝗮𝗹 𝗙𝗲𝗮𝘁𝘂𝗿𝗲 𝗘𝘅𝘁𝗿𝗮𝗰𝘁𝗶𝗼𝗻 (𝗣𝘆𝗧𝗼𝗿𝗰𝗵) A standard convolutional neural network processes the 28×28 MNIST images and extracts high-level features. This step reduces the dimensionality of the input while preserving essential information. 𝗙𝗲𝗮𝘁𝘂𝗿𝗲 𝗛𝗮𝗻𝗱𝗼𝗳𝗳 The CNN outputs two features, chosen to match the number of qubits used in the quantum circuit. 𝗤𝘂𝗮𝗻𝘁𝘂𝗺 𝗣𝗿𝗼𝗰𝗲𝘀𝘀𝗶𝗻𝗴 𝗟𝗮𝘆𝗲𝗿 (𝗤𝗶𝘀𝗸𝗶𝘁) The features are encoded into a quantum state using a ZZFeatureMap. A parameterized variational circuit (RealAmplitudes) transforms this state, and the expectation value of a measurement operator is used for classification. 𝗘𝗻𝗱-𝘁𝗼-𝗘𝗻𝗱 𝗧𝗿𝗮𝗶𝗻𝗶𝗻𝗴 Using Qiskit’s TorchConnector, the quantum and classical components are trained jointly. Gradients from the quantum circuit are computed using the parameter-shift rule and integrated with PyTorch’s automatic differentiation. 𝗧𝗼𝗼𝗹𝘀 𝗮𝗻𝗱 𝗦𝗲𝘁𝘂𝗽 • PyTorch and Qiskit Machine Learning • Training performed on the Aer simulator, with compatibility for quantum hardware via primitives On small subsets of the dataset, this hybrid approach achieves perfect classification accuracy, demonstrating how classical feature extraction combined with quantum decision layers can be effective even with limited qubit resources. Hybrid quantum–classical models provide a practical direction for near-term quantum machine learning, especially when classical preprocessing reduces problem complexity before quantum evaluation. 𝗥𝗲𝗳𝗲𝗿𝗲𝗻𝗰𝗲𝘀 𝗮𝗻𝗱 𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴 𝗥𝗲𝘀𝗼𝘂𝗿𝗰𝗲𝘀 https://lnkd.in/gkwjwp-r https://lnkd.in/g7ge8MHD https://lnkd.in/gApuMGHE ♻️ Repost if you found this valuable! ➕ Follow me https://lnkd.in/gGhxx66A for more insights on the cutting edge of AI and Quantum. #QuantumComputing #QuantumMachineLearning #MachineLearning #PyTorch #Qiskit #ArtificialIntelligence #HybridModels