Quantum Feature Engineering for Credit Default Prediction: When and Why IQP Circuits Help Linear Classifiers (arXiv:2609.10505)

Forum topic · 小凯 · 2026-09-11

Summary

A paper on arXiv (2609.10505) by Finkelstein, Levy, Yakhini, and Cohen investigates whether Instantaneous Quantum Polynomial-time (IQP) circuits can generate features that improve credit default prediction. Using the UCI Default of Credit Card Clients dataset with five-fold cross-validation, the authors select 8 of 23 financial attributes, encode them as rotation angles in an 8-qubit IQP circuit, and extract 16 expectation values as new features. Appending these quantum features to Logistic Regression raises F1 from 0.462 to 0.517 (+0.055, p < 0.0001), while Kernel PCA at the same feature budget only reaches 0.493; the gap survives Benjamini-Hochberg correction across 12 tests (p = 0.00007). Notably, Random Forest, SVM, XGBoost, and k-NN show no benefit, pointing to a linear-expressivity mechanism rather than a generic improvement. Feature selection also matters: Random Forest importance-guided selection achieves F1 = 0.523, while encoding uncorrelated features drops performance to 0.496, suggesting the circuit amplifies existing informative structure rather than creating it.

Paper Overview

Field: Machine Learning Authors: Menachem Finkelstein, Diana Legziel Levy, Zohar Yakhini, Sarel Cohen Published: 2026-09-09 arXiv: 2609.10505

Abstract

Credit default prediction is a tabular classification problem in which modest gains in F1 translate directly into reduced financial exposure. We ask whether Instantaneous Quantum Polynomial-time (IQP) circuits can produce features that improve a classifier over both its raw classical baseline and Kernel PCA - the strongest unsupervised classical non-linear alternative - at an equal feature budget. The dataset provides 23 financial attributes per client; for an n-qubit circuit we select n of them, encode each as a rotation angle, and read 2n expectation values back out as new features. The motivation for using a quantum circuit is computational: an n-qubit IQP circuit runs in constant depth and encodes feature correlations in a 2^n-dimensional Hilbert space, whereas classical simulation of its exact output statistics scales exponentially in n. Using the UCI Default of Credit Card Clients dataset and five-fold cross-validation, we find that appending 16 IQP features (n = 8 qubits) to a Logistic Regression model raises F1 from 0.462 to 0.517 (+0.055, p < 0.0001). Kernel PCA, the next-best method, reaches only 0.493 at the same feature count; the gap survives Benjamini-Hochberg correction across 12 tests (p = 0.00007). No other classifier - Random Forest, SVM, XGBoost, or k-NN - benefits, which points to a linear-expressivity mechanism rather than a generic improvement. We also show that how the 8 input features are chosen matters: Random Forest importance-guided selection reaches F1 = 0.523, while encoding maximally uncorrelated features drops it to 0.496, demonstrating that the circuit amplifies informative structure rather than creating it from scratch.

Key Findings

Main result: 16 IQP features (8 qubits) appended to Logistic Regression improve F1 from 0.462 to 0.517 (+0.055, p < 0.0001) on the UCI Default of Credit Card Clients dataset with five-fold cross-validation.
Classical comparison: Kernel PCA, the strongest unsupervised classical non-linear alternative, reaches only F1 = 0.493 at the same feature budget; the gap survives Benjamini-Hochberg correction across 12 tests (p = 0.00007).
Mechanism: Only linear classifiers benefit — Random Forest, SVM, XGBoost, and k-NN show no improvement — suggesting a linear-expressivity mechanism rather than a generic performance boost.
Feature selection matters: Random Forest importance-guided selection achieves F1 = 0.523, while encoding maximally uncorrelated features drops performance to 0.496, indicating the circuit amplifies existing informative structure rather than creating it.
Computational motivation: An n-qubit IQP circuit runs in constant depth and encodes feature correlations in a 2^n-dimensional Hilbert space, while exact classical simulation scales exponentially in n.

--- *Auto-collected on 2026-09-11*