Paper: arXiv 2502.01495

Authors: Joshua Rosaler, Luca Candelori, Vahagn Kirakosyan, Kharen Musaelian, Ryan Samson, Martin T. Wells, Dhagash Mehta, Stefano Pasquali

Abstract

We investigate the application of quantum cognition machine learning (QCML), a novel paradigm for both supervised and unsupervised learning tasks rooted in the mathematical formalism of quantum theory, to distance metric learning in corporate bond markets. Compared to equities, corporate bonds are relatively illiquid and both trade and quote data in these securities are relatively sparse. Thus, a measure of distance/similarity among corporate bonds is particularly useful for a variety of practical applications in the trading of illiquid bonds, including the identification of similar tradable alternatives, pricing securities with relatively few recent quotes or trades, and explaining the predictions and performance of ML models based on their training data. Previous research has explored supervised similarity learning based on classical tree-based models in this context; here, we explore the application of the QCML paradigm for supervised distance metric learning in the same context, showing that it outperforms classical tree-based models in high-yield (HY) markets, while giving comparable or better performance (depending on the evaluation metric) in investment grade (IG) markets.

Complexity vs Empirical Score

  • Math Complexity: 8.5/10
  • Empirical Rigor: 8.0/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper introduces advanced quantum mathematics (Hilbert spaces, operators, fidelity) for bond similarity, indicating high mathematical complexity. It includes a detailed empirical methodology with backtesting on real bond data and quantitative performance comparisons against benchmarks, demonstrating high empirical rigor.

Research Flowchart

  flowchart TD
  A["Research Goal<br>Develop supervised similarity<br>for illiquid corporate bonds<br>using Quantum Cognition ML (QCML)"] --> B["Input Data<br>High-Yield & Investment Grade<br>Corporate Bond Data"]
  B --> C["Methodology<br>Compare QCML vs.<br>Classical Tree-Based Models"]
  C --> D["Process<br>Train Models & Learn<br>Distance Metrics"]
  D --> E["Evaluation<br>Performance on Trading &<br>Pricing Tasks"]
  E --> F{"Key Findings & Outcomes"}
  F --> G["QCML Outperforms<br>Classical Models in<br>High-Yield Markets"]
  F --> H["QCML Achieves Comparable/<br>Better Performance in<br>Investment Grade Markets"]

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 applies quantum cognition machine learning (QCML) to supervised distance metric learning for corporate bonds, showing it outperforms classical tree‑based models in high‑yield markets.

Source sentence
we explore the application of the QCML paradigm for supervised distance metric learning in the same context, showing that it outperforms classical tree-based models in high-yield (HY) markets

Assumes 5

  • Credit spread is used as the prediction target for investment‑grade bonds (IGSB). data
    Source sentence
    We choose credit spread, the difference between the bond’s yield and the yield of a treasury bond of comparable maturity, as the prediction target for IGSB
  • Bond yield is used as the prediction target for high‑yield bonds (HYG). data
    Source sentence
    We choose the bond yield (return on initial investment up to time of maturity) rather than spread as the prediction target for HYG bonds
  • Weights for KNN averaging should fall off with distance (or increase with proximity). model
    Source sentence
    Apart from the general assumption that they should generally fall off with distance d (or equivalently, increase with proximity), the question of which weights one should use is largely subjective and ill-defined
  • QCML creates a global quantum manifold that is robust to noise. model
    Source sentence
    It creates a global quantum manifold model (in the sense of quantum geometry ...)
Show 1 more
  • Corporate bond data are relatively illiquid and sparse. market
    Source sentence
    Compared to equities, corporate bonds are relatively illiquid and both trade and quote data in these securities are relatively sparse

Data

  • HYG corporate bond dataset market-prices unknown
    high‑yield corporate bonds
    Source sentence
    For the corporate bonds dataset we see in Figure 4 that for HYG, the QCML metric outperforms other metrics
  • IGSB corporate bond dataset market-prices unknown
    investment‑grade corporate bonds
    Source sentence
    For IGSB, we see that the QCML metric outperforms again for the unweighted KNN regression
  • California Housing benchmark public
    Source sentence
    Among the public datasets, we see that for the California Housing dataset, where QCML and RF perform on par
  • Student Performance benchmark public
    Source sentence
    On the Student Performance dataset, where the QCML regressor performs better than or on par with RF and Linear Regression
  • Diabetes benchmark public
    Source sentence
    In the Diabetes dataset, where QCML under-performs as a regression model, it performs roughly on par

Costs

Not applicable The paper reports no trading or execution result.

Finds 3

  • In KNN regression on HYG bonds, the QCML distance metric outperforms Euclidean and RF‑GAP metrics for both unweighted and proximity‑weighted methods. performance
    Source sentence
    For the corporate bonds dataset we see in Figure 4 that for HYG, the QCML metric outperforms other metrics for both the unweighted and proximity-weighted methods of KNN regression
  • On public datasets, QCML’s metric generally outperforms Euclidean and RF‑GAP in unweighted KNN, and is competitive in proximity‑weighted KNN. performance
    Source sentence
    For public datasets, we see in Figure 5 that the QCML metric again outperforms both RF-GAP and Eu‑clidean metrics using the unweighted method of KNN regression
  • Linear regression outperforms both Random Forest and QCML on the Diabetes dataset, highlighting that QCML is not uniformly superior. negative
    Source sentence
    It is interesting to note that linear regression works well for the Diabetes dataset, outperforming both RF

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 36 related papers). Coverage is partial, so this is neither replication nor refutation.

Cited by 1 paper in the archive (1 by independent authors)

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 (33)

  1. Introduction to Latent Variable Energy-Based Models: A Path Towards Autonomous Machine Intelligence (2022) — link
  2. Quantum Models of Cognition and Decision (2012) — link
  3. Quantum Cognition (2022) — link
  4. Applications of Quantum Statistics in Psychological Studies of Decision Processes (1995) — link
  5. Quantum-like brain: “Interference of minds” (2006) — link
  6. Quantum (matrix) geometry and quasi-coherent states (2021) — link
  7. Matrix geometry and coherent states (2015) — link
  8. Quantum Cognition Machine Learning: AI Needs Quantum (2024)
and 25 more
  1. Robust estimation of the intrinsic dimension of data sets with quantum cognition machine learning (2024)
  2. Quantum Cognition Machine Learning: financial forecasting (2024)
  3. The Interaction of Value and Momentum Strategies (1997)
  4. Empirical Asset Pricing via Machine Learning (2020) — link
  5. Common risk factors in the returns on stocks and bonds (1993) — link
  6. On Persistence in Mutual Fund Performance (1997) — link
  7. Accrual reliability, earnings persistence and stock prices (2005) — link
  8. Replicating Anomalies (2020) — link
  9. Random Forests (2001)
  10. Random forests (2009)
  11. Geometry- and Accuracy-Preserving Random Forest Proximities (2023) — link
  12. Random forests and adaptive nearest neighbors (2006)
  13. Classification and regression by randomForest (2002)
  14. Quantifying Outlierness of Funds from their Categories using Supervised Similarity — → 2308.06882
  15. Supervised similarity learning for corporate bonds using Random Forest proximities (2022)
  16. Enhanced Local Explainability and Trust Scores with Random Forest Proximities — → 2310.12428
  17. Open Set Recognition for Random Forest (2024) — link
  18. Can an unsupervised clustering algorithm reproduce a categorization system? — → 2408.10340
  19. Quantile Regression using Random Forest Proximities — → 2408.02355
  20. Case-based Explainability for Random Forest: Prototypes, Critics, Counter-factuals and Semi-factuals — → 2408.06679
  21. Machine Learning-based Relative Valuation of Municipal Bonds — → 2408.02273
  22. Quantum Geometry, Matrix Theory, and Gravity (2024)
  23. RF-GAP Python: Python code for Random Forest-Geometry- and Accuracy-Preserving Proximities (2024)
  24. rfproximity: A Python package for RF proximity analysis (2024)
  25. Quantum Computation and Quantum Information (2000)

6 of 33 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.