Discover, Connect & Collaborate at TECHINNOVATION 2021

Quantum Binary Classifiers for Noisy Datasets

Technology Overview

This technology offer is a quantum machine learning algorithm applied to binary classification models for noisy datasets which are prevalent in financial and other datasets. By combining hybrid-neural networks, quantum parametric circuits, and data re-uploading we have improved the classification of non-convex 2-dimensional figures by understanding learning stability as noise increases in the dataset. The metric we use for assessing the performance of our quantum classifiers is the area under the receiver operator curve (ROC AUC). We are interested to collaborate with partners with use cases for binary classification of noisy data. Also, as quantum technology is still insufficient for large datasets, we would be interested to work with technology partners for assessing implementation paths.

Technology Features, Specifications and Advantages

The innovation is to combine a hybrid of quantum and classical classifiers to provide improved performance over simple classical and quantum classifiers for noisy datasets. These can be used to simultaneously increase the true positive (TP) and reduce false positives (FP) for example adding value for lenders requiring to decide on whether a borrower will default on a loan or not. Increasing TP provides greater returns from more productive lending and decreasing FP creates more access to funding. Quantum technology is new and will continue to improve over time further enabling returns. We see improvements of more than 0.05 in the difference in ROC AUC across noise levels which especially increases further at high noise levels.

Potential Applications

The primary application area this is being developed for is credit rating. However, anywhere that uses binary classification for noisy datasets may also benefit. For example:

  • Predicting whether a skin lesion is benign or malignant based on some criteria.
  • Understanding consumer behaviour websites leading to product purchases.
  • Providing a decision framework for screening new job candidates.
  • Detecting anomalous and fraudulent activity in banking transactions.

Products targeting the above use cases could be developed when quantum technology is sufficiently mature. For credit rating alone the estimated size of the market for this technology is in the billions of dollars.

Furthermore, the methodology used to combine classical, variational and data-reuploading could be applied beyond binary classification into other areas of machine learning such as portfolio optimization and pricing derivatives.

Customer Benefits

Many customers could benefit the further application of this technology. In credit ratings the underbanked could be better served providing access to loans with better interest rates as well as improving improved risk management for finance providers with costs saved from bad debt. Financial systems and economies are also better served with improved credit facilities.

Contact Person

Francis Lim


Singapore Management University - School of Computing and Information Systems

Technology Category

  • Infocomm
  • Artificial Intelligence, Computer Simulation & Modeling, eCommerce & ePayment, Financial Technology

Technology Readiness Level


AI, ML, machine learning, quantum, binary classification