We present a simple and efficient way to reduce the contraction cost of a tensor network to simulate a quantum circuit. We start by interpreting the circuit as a ZX-diagram. We then use simplification and local complementation rules to sparsify it. We find that optimizing graph-like ZX-diagrams improves existing state of the art contraction cost by several order of magnitude. In particular, we demonstrate an average contraction cost 1180 times better for Sycamore circuits of depth 20, and up to 4200 times better at peak performance.

翻译：我们提出了一种简单高效的方法，通过降低张量网络的收缩成本来模拟量子电路。首先将电路解释为ZX图，随后应用简化规则与局部互补规则对其进行稀疏化处理。研究发现，优化类图ZX图可使现有最优收缩成本降低数个数量级。具体而言，对于深度为20的Sycamore电路，我们实现了平均收缩成本降低1180倍，峰值性能下可达4200倍提升。