# Techniques: A Brief Review of Quantum Machine Learning for Financial Services

## What the paper contributes (one paragraph)
This is a focused review paper, not a method paper. It surveys the state of the art in quantum machine learning and maps it onto financial services, concentrating on the regime of quantum algorithms over classical data because that is what aligns with finance. It first recaps how classical machine learning is used for credit scoring, risk management and compliance, fraud detection, stock price prediction, and personalisation, then gives an opinionated, deliberately hype-free assessment of the general promises and limitations of QML. It then walks through three families of QML methods drawn from the literature, supervised learning with quantum-enhanced feature spaces (Quantum Variational Classifier, Quantum Kernel Estimation, Quantum Neural Networks), quantum generative AI (Quantum Transformers, Quantum Circuit Born Machines, Quantum Boltzmann Machines, Quantum GANs), and Quantum Graph Neural Networks, each tied to the finance applications most likely to benefit. Every concrete algorithm it describes is attributed by citation to prior work; the paper itself originates no new algorithm, bound, or protocol.

## Techniques offered
- **QML-to-finance survey and taxonomy with an anti-hype promises-and-limitations assessment** - curates and organises existing QML methods (variational classifiers, quantum kernels, QNNs, quantum transformers, QGNNs, QCBM/QBM/QGAN) into three families and maps each to financial use cases (credit scoring, risk management, fraud detection, stock price prediction), supplying a structured, citation-backed landscape and a sober promises-versus-limitations checklist rather than a new capability.
  - guarantee: none; a review supplies an organised, referenced overview and qualitative judgements, not a statistical bound, soundness property, certificate, or static proof.
  - quote: "We also provide an overview of the challenges, potential, and limitations of QML, both in these specific areas and more broadly across the field."

## Where it could apply
- Target primitive(s): VQE / variational quantum classifiers, quantum kernel estimation, and other parameterised-quantum-circuit (QNN) routines as discussed targets, but none supplied as a reusable primitive.
- Target application group(s): quantum_ml; finance.
- Code family / hardware assumptions: NISQ gate-model devices are the near-term framing, with a neutral atom processor and quantum annealers noted for specific cited finance results; fault-tolerant quantum computing flagged only as a longer-term horizon. No code family is assumed by the paper itself.

## Caveats
This is a literature review and contains nothing suppliable to an external vendor or application result in the technique sense the matrix wants. Every algorithm it describes (Quantum Variational Classifier, Quantum Kernel Estimation, QNNs, the three Quantum Transformer variants, QCBM, QBM, QGAN, QGNN) is attributed by citation to other papers, and the provable-advantage claims it relays (for example the PROMISEBQP-complete separation and the quantum-SVM speedup) belong to those cited works, not to this one. The single bullet above is offered as the best fit, namely the curated survey and taxonomy itself, which carries no formal guarantee. Full LaTeX source was available; no paywall or partial access. Do not use the em dash character.