.jpg)
Quantum computing is advancing toward industrial applications, but choosing the right paradigm remains a challenge. A recent study compared, under controlled conditions, two predominant approaches: continuous variable (CV) and discrete variable (DV), applied to defect classification in semiconductor wafers (WM-811K dataset). To isolate the impact of the quantum circuit, a common convolutional network of ~4.3 million parameters was used, feeding interchangeable heads: one classical, one CV-QNN, and one DV-QNN, each scaled to 3, 4, and 8 qumodes/qubits. The results show that the CV head clearly outperforms the DV: with 4 qumodes it achieves 79.7% accuracy versus 61.6% for DV, an 18-point gap. The advantage is especially pronounced in the Edge-Loc class, where CV achieves a recall of 0.66 while DV does not exceed 0.05, demonstrating that the structured CV layer better captures fine spatial distinctions. Training curves reveal that DV's limitation is a representational capacity ceiling, not an optimization failure. Although both quantum heads fall below the classical baseline (85.0%), the study isolates where a structured head already adds value and how, with improvements in noise and scale, each paradigm could offer practical advantages.
For companies exploring these technologies, having custom applications is key to integrating quantum or classical solutions into real production workflows. At Q2BSTUDIO we develop custom software that combines artificial intelligence, AI agents, and AWS and Azure cloud services to optimize processes such as quality control in manufacturing. Additionally, we offer business intelligence services with Power BI and cybersecurity to protect critical data. Defect classification is just one example of how AI for businesses can transform industry, and our team helps implement these capabilities through solutions tailored to each need.
