New quantum-inspired tensor network algorithm SMT-AD offers efficient, scalable anomaly detection with linear parameter growth

SMT-AD uses superposition of bond-dimension-1 matrix product operators combined with Fourier-assisted feature embedding for anomaly detection

The approach is highly parallelizable with learnable parameters that scale linearly with feature size, embedding resolutions, and matrix product operator components

Demonstrates competitive performance against established baselines on standard datasets including credit card transaction anomaly detection

Achieves strong results even with minimal configurations, offering practical advantages for real-world applications