Unisys research accepted at IEEE Quantum Artificial Intelligence 2025

Quantum computing is advancing quickly, offering enterprises computational capabilities beyond traditional systems. Combined with AI, it can address complex optimization problems and strengthen critical business applications.

Demonstrating progress in this field, two Unisys research papers have earned acceptance at the 2025 IEEE International Conference on Quantum Artificial Intelligence following rigorous peer review. This acceptance confirms the technical merit and innovation of both research contributions.

The two papers address different but complementary aspects of quantum computing advancement:

Paper 1: Infrastructure breakthrough eliminates quantum embedding delays

Quantum computers need efficient methods to map complex business problems onto their hardware. Existing approaches create long, unstable connections that break frequently, causing errors and limiting problem sizes.

Unisys research "Optimized Quantum Embedding: A Universal Minor-Embedding Framework for Large Complete Bipartite Graph" developed a universal framework that eliminates these unstable connections while dramatically reducing processing time.

Performance gains delivered:

• 99.98% reduction in embedding time—from hours to milliseconds
• Complete elimination of unstable connections that cause failures
• Universal compatibility across quantum hardware architectures
• Enhanced stability enabling larger, more complex problem processing

This advancement opens quantum computing to supply chain optimization, logistics planning, and financial portfolio management where processing speed determines competitive advantage.

Paper 2: Quantum-enhanced AI detects financial risks traditional systems miss

Financial institutions struggle to identify potential loan defaults in severely imbalanced datasets where defaults represent fewer than 5% of cases. Traditional AI systems miss these rare but costly events.

"Enhancing Credit Risk Prediction through Co-trained Hybrid Quantum Transfer Learning Model" research by team members Salvatore Sinno and Shruthi Thuravakkath from Unisys Enterprise Computing Solutions, alongside academic collaborators from IIIT Kottayam, developed a hybrid quantum-classical model that simultaneously trains both components. This co-training approach enables AI systems to better recognize minority cases.

Results achieved:

• 91.17% accuracy in credit risk classification on real-world datasets
• 73% recall for minority class detection—identifying potential defaults with unprecedented sensitivity
• 0.9134 AUC score demonstrating superior reliability in risk ranking
• End-to-end optimization where both components adapt together

This quantum-enhanced approach helps financial institutions identify risks that traditional systems miss.

Advancing quantum computing from research to enterprise reality

The IEEE International Conference on Quantum Artificial Intelligence brings together leading researchers and industry practitioners to advance quantum computing applications.

To explore quantum computing applications, connect with Unisys quantum computing experts to discuss how these advances might benefit your organization.