Reproduce “Quantum Scrambling Born Machine” on OpenResearch

About this paper

Marcin Płodzień

Researchers from Qilimanjaro Quantum Tech introduced the Quantum Scrambling Born Machine (QSBM), a quantum generative model designed for near-term quantum processors that leverages a fixed scrambling unitary and trainable single-qubit rotations. The QSBM achieved competitive performance against classical generative models on complex 2D distributions with matched small parameter counts, and demonstrated enhanced expressivity through the use of ancilla qubits.

Key insights

• Quantum scrambling serves as a robust and generic entangling resource; the QSBM's generative performance primarily depends on achieving Haar-typical entanglement, showing weak sensitivity to the specific microscopic implementation of the scrambler (e.g., Haar-random, brickwork circuit, or analog Hamiltonian).
• Decoupling entanglement from trainable parameters leads to a hardware-efficient and parameter-efficient quantum generative model, making it suitable for near-term quantum devices.
• Tracing out ancilla qubits genuinely boosts the expressivity of the Born machine by enabling the generation of mixed-state distributions, which can better fit complex target distributions.

Summary by alphaXiv.