New quantum machine learning framework breaks 16-qubit entanglement barrier
Researchers at the Karlsruhe Institute of Technology and the Technical University of Applied Sciences Regensburg have created a new framework for quantum machine learning. The system uses advanced diagnostics and pulse-level optimisation to improve how quantum models perform. It marks a shift from older gate-based methods by directly controlling qubit operations. The framework relies on Fourier-analytic tools and expanded entanglement metrics to assess quantum circuits. Unlike traditional approaches, it optimises control pulses applied to qubits, allowing for more flexible and efficient designs. This method also supports composable ansatz constructions, enabling end-to-end tuning of pulse parameters.
A key breakthrough is its ability to analyse entanglement in systems of up to 16 qubits—six times larger than previous limits. This expansion is particularly useful for quantum error correction and fault-tolerant computing, where larger qubit arrays are essential. The framework also introduces tailored error mitigation, adapting strategies to the constraints of modern quantum hardware.
Built with the JAX library and a dedicated quantum simulator, the system ensures scalability and high performance. By moving beyond gate-based models, it opens new possibilities for optimising quantum algorithms directly at the pulse level. The new framework raises practical questions about balancing complexity and real-world use in quantum computing. Its ability to handle 16-qubit entanglement and pulse-level optimisation provides a foundation for more robust quantum applications. Developers can now explore more efficient error correction and circuit designs for current and future hardware.
