IBM Quantum reposted this
As we improve our quantum hardware, classical tensor network algorithms are also evolving continuously. A new preprint from IBM researchers, in collaboration with researchers from AMD and Basque Quantum (BasQ), considers how state-of-the-art tensor network techniques can be used to enhance a quantum workflow, such that progress on the classical side can be translated into progress for quantum computing. The researchers introduce a new error mitigation technique called Lindblad Multiproduct Formulas (https://lnkd.in/efSR7g2C). After learning the structure of the noise on the quantum computer, the technique uses two-dimensional tensor networks contracted with loop-corrected belief propagation to mitigate the errors in the expectation values obtained from the quantum computer. The full workflow is applied to a model of two-dimensional discrete time crystals (https://lnkd.in/eN9T3k59) which was studied in work by BasQ and IBM earlier this year. The researchers found that running the classical part of this algorithm on an AMD GPU resulted in a 5x speedup. In recent months, we also demonstrated how classical simulation methods such as Pauli propagation can be employed in more efficient error mitigation workflows (https://lnkd.in/esqq5G-Q | https://lnkd.in/eqjzDT9N). These works highlight the growing synergy that can arise from classical and quantum computers working together, that will enable computations that are beyond their individual capabilities.