Paper: arXiv 2404.00015

Authors: Javier Mancilla, André Sequeira, Tomas Tagliani, Francisco Llaneza, Claudio Beiza

Abstract

Quantum Kernels are projected to provide early-stage usefulness for quantum machine learning. However, highly sophisticated classical models are hard to surpass without losing interpretability, particularly when vast datasets can be exploited. Nonetheless, classical models struggle once data is scarce and skewed. Quantum feature spaces are projected to find better links between data features and the target class to be predicted even in such challenging scenarios and most importantly, enhanced generalization capabilities. In this work, we propose a novel approach called Systemic Quantum Score (SQS) and provide preliminary results indicating potential advantage over purely classical models in a production grade use case for the Finance sector. SQS shows in our specific study an increased capacity to extract patterns out of fewer data points as well as improved performance over data-hungry algorithms such as XGBoost, providing advantage in a competitive market as it is the FinTech and Neobank regime.

Complexity vs Empirical Score

  • Math Complexity: 7.5/10
  • Empirical Rigor: 4.0/10
  • Quadrant: Lab Rats — high math complexity, limited empirical testing

Why this score: The paper introduces advanced quantum machine learning concepts like quantum kernels, Hilbert space mappings, and evolutionary algorithms for feature map optimization, requiring sophisticated mathematical background, but the empirical validation is limited to preliminary results on a single production dataset without reported statistical significance or backtest metrics.

Research Flowchart

  flowchart TD
  A["Research Goal: Enhance Credit Scoring<br>in FinTech with QML"] --> B["Methodology: Systemic Quantum Score SQS"]
  C["Data: Scarce/Skewed Financial Datasets"] --> B
  B --> D["Computational Process:<br>Quantum Kernels for Feature Space"]
  D --> E{"Classical Models vs SQS"}
  E -- Classical (e.g., XGBoost) --> F["Struggles with Data Scarcity"]
  E -- SQS Proposed --> G["Key Outcomes:<br>Increased Pattern Extraction<br>Improved Generalization"]

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper proposes a novel method called Systemic Quantum Score (SQS) that uses evolutionary algorithms to design efficient quantum kernels for credit scoring.

Source sentence
We introduce a novel method named Systemic Quantum Score (SQS) that leverage evolutionary algorithms [3] for efficient Quantum Kernel design.

Assumes 4

  • Quantum kernels can generalize better than classical models on scarce and unbalanced datasets. data
    Source sentence
    Quantum feature spaces are projected to find better links between data features and the target class to be predicted even in such challenging scenarios
  • Instantaneous Quantum Polynomial (IQP) embeddings cannot be easily replicated by classical devices. model
    Source sentence
    it is essential to ensure that these feature maps cannot be easily replicated by classical devices, lest the quantum device’s unique advantage be nullified
  • Evolutionary algorithms can efficiently search the space of quantum feature maps. model
    Source sentence
    Gradient-free optimization strategies such as evolutionary algorithms provide the ability to scout part of the potential space empowering at the same time improvements from generation to generation
  • Lower‑qubit quantum maps are sufficient for practical performance. model
    Source sentence
    In order to make it more tractable lower dimensional qubit mappings were selected in moving forward

Data

  • Fintonic loan and fraud dataset · Fintonic alternative proprietary
    Source sentence
    One of Fintonic’s most valuable assets lies in its extensive data reservoir.

Costs

Not applicable The paper reports no trading or execution result.

Finds 3

  • SQS outperforms both classical SVC and XGBoost models in the down‑sampled regime (500–3000 samples), though XGBoost catches up on the full dataset. performance
    Source sentence
    This approach was able to outperform both classical models (SVC and XGBoost) in the down‑sampled regime between 500 and 3000 samples where XGBoost started to catch up outperforming finally for the whole dataset.
  • Increasing the number of qubits does not significantly improve performance; fitness gains stem from diversity in the initial population. performance
    Source sentence
    Althought, no significant improvement was found on the scaling of the kernels as shown in table 2, being of similar performance in all qubit ranges tested up to the maximum 10 features
  • Lower‑qubit feature maps are preferred for hardware feasibility without loss of discriminatory power. other
    Source sentence
    In order to be able to be as pragmatic as possible and look into potential future works performing on hardware experiments, we prioritized the usage of lower qubit-count maps.

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 14 related papers). Coverage is partial, so this is neither replication nor refutation.

Code

No repository by the authors found in the paper.

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.

Cited References (50)

  1. A Structured Survey of Quantum Computing for the Financial Industry (2022) — link
  2. On quantum backpropagation, information reuse, and cheating measurement collapse (2024)
  3. Better than classical? The subtle art of benchmarking quantum machine learning models (2024) — link
  4. A Preprocessing Perspective for Quantum Machine Learning Classification Advantage in Finance Using NISQ Algorithms (2022) — link
  5. Quantum Kernels for Real-World Predictions Based on Electronic Health Records (2022) — link
  6. Kernel Alignment for Quantum Support Vector Machines Using Genetic Algorithms (2023) — link
  7. Quantum support vector machine for big data classification (2014)
  8. Quantum Multiple Kernel Learning (2020) — link
and 42 more
  1. Kernel principal component analysis (1997)
  2. Quantum-Classical Multiple Kernel Learning (2023) — link
  3. Efficient Parameter Optimisation for Quantum Kernel Alignment: A Sub-sampling Approach in Variational Training (2024) — link
  4. Automatic design of quantum feature maps (2021) — link
  5. Analysis and synthesis of feature map for kernel-based quantum classifier (2020) — link
  6. Deep feature synthesis: Towards automating data science endeavors (2015)
  7. XGBoost: A Scalable Tree Boosting System (2016) — link
  8. O ptuna: A Next-Generation Hyperparameter Optimization Framework (2019) — link
  9. Quantum Multiple Kernel Learning in Financial Classification Tasks (2023) — link
  10. Multiple Kernel Learning Algorithms (2011)
  11. Power Characterization of Noisy Quantum Kernels (2024) — link
  12. Quantum Machine Learning Applied to the Classification of Diabetes (2022) — link
  13. Unsupervised quantum machine learning for fraud detection (2022) — link
  14. Towards understanding the power of quantum kernels in the NISQ era (2021) — link
  15. Quantum machine learning beyond kernel methods (2023) — link
  16. Quantum machine learning for finance ICCAD Special Session Paper (2021)
  17. Default of credit card clients (2016)
  18. An introduction to ROC analysis (2006) — link
  19. Quantum dimensionality reduction by linear discriminant analysis (2023) — link
  20. Dimensionality Reduction of Multimodal Labeled Data by Local Fisher Discriminant Analysis (2007)
  21. Quantum Embedding Search for Quantum Machine Learning (2022) — link
  22. Power of data in quantum machine learning (2021) — link
  23. Training quantum embedding kernels on near-term quantum computers (2022) — link
  24. A survey on deep learning tools dealing with data scarcity: definitions, challenges, solutions, tips, and applications (2023)
  25. The role of entanglement for enhancing the efficiency of quantum kernels towards classification (2023) — link
  26. Quantum Machine Learning in Feature Hilbert Spaces (2019) — link
  27. Exponential concentration and untrainability in quantum kernel methods (2022) — link
  28. Data re-uploading for a universal quantum classifier (2020) — link
  29. Average-Case Complexity Versus Approximate Simulation of Commuting Quantum Computations (2016) — link
  30. Quantum computing for finance (2023) — link
  31. A didactic approach to quantum machine learning with a single qubit (2023) — link
  32. Importance of kernel bandwidth in quantum machine learning (2022) — link
  33. Generating quantum feature maps for SVM classifier (2022) — link
  34. Several fitness functions and entanglement gates in quantum kernel generation (2023) — link
  35. Least Squares Support Vector Machine Classifiers (1999) — link
  36. Quantum Principal Component Analysis Only Achieves an Exponential Speedup Because of Its State Preparation Assumptions. (2018) — link
  37. Learning With Kernels: Support Vector Machines, Regularization, Optimization, and Beyond (2001) — link
  38. Quantum machine learning models are kernel methods — link
  39. Supervised learning with quantum-enhanced feature spaces (2018) — link
  40. Structure learning of quantum embeddings (2022) — link
  41. Foundations of Machine Learning (2018)
  42. Generalization in quantum machine learning from few training data (2022) — link

0 of 50 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.