ML-Powered FPGA-based Real-Time Quantum State Discrimination Enabling Mid-circuit Measurements

Similar to reading the transistor state in classical computers, identifying the quantum bit (qubit) state is a fundamental operation to translate quantum information. However, identifying the qubit state has been the slowest and most susceptible to errors operation on superconducting quantum processors. Most existing qubit state discriminating algorithms have only been implemented and optimized “after the fact”—using offline data transferred from a quantum control circuit to host computers. Real-time state discrimination is not possible because a superconducting qubit state only survives for a few hundred μs, which is much shorter than the communication delay between the readout circuit and the host computer (i.e., tens of ms). Mid-circuit measurement, where measurements are conducted on qubits at intermediate stages within a quantum circuit rather than solely at the end, represents an advanced technique for qubit reuse in quantum computing. This approach expands the computational toolkit, enabling the implementation of more sophisticated error correction algorithms and maximizing the potential of the noisy intermediatescale quantum (NISQ) era devices available today. For midcircuit measurements necessitating single-shot readout, it is imperative to employ an in-situ technique for state discrimination characterized by low latency and high accuracy. This paper introduces QubiCML, a field-programmable gate array (FPGA) based system for real-time qubit state discrimination enabling mid-circuit measurement—the ability to measure the qubit state at the electronic control circuit before/without transferring quantum data to a host computer. QubiCML provides in-situ real-time feedback and verification for quantum algorithm development and optimization. A multi-layer neural network has been designed and deployed on the FPGA platform to ensure accurate in-situ state discrimination. For the first time, ML-powered quantum state discrimination has been implemented on a radio frequency system-on-chip (RFSoC) FPGA platform (Xilinx ZCU216). The deployed lightweight network on the FPGA hardware only takes 54 ns to complete each inference (state measurement). We evaluated QubiCML’s performance on in-house superconducting quantum processors and obtained an average accuracy of 98.5% with only 500 ns readout length. QubiCML has the potential to become the standard real-time state discrimination method for the quantum community