Fig. 2: Quantum circuit classifier with 18 superconducting qubits for learning three sequential tasks.
From: Experimental demonstration of quantum continual learning with superconducting qubits
The circuit consists of four blocks of operations with a total of 216 variational parameters. Each block performs three consecutive single-qubit rotation gates on all qubits, followed by two layers of CNOT gates applied to adjacent qubits. The quantum classifier adapts the interleaved block encoding strategy to encode classical data and naturally handles the quantum data (in the form of quantum states) as input. For each input data, the classifier determines the prediction label based on the local observable \(\langle {\widehat{\sigma }}_{9}^{z}\rangle\): label 0 and label 1 for \(\langle {\widehat{\sigma }}_{9}^{z}\rangle \ge 0\) and \(\langle {\widehat{\sigma }}_{9}^{z}\rangle < 0\), respectively.
