Quantum reservoir computing induced by controllable damping

Quantum reservoir computing has emerged as a promising machine learning paradigm for processing temporal data on near-term quantum devices, as it allows for exploiting the large computational capacity of the qubits without suffering from typical issues that occur when training a variational quantum circuit. In particular, quantum gate-based echo state networks have proven effective for learning when the evolution of the reservoir circuit is non-unital. Nonetheless, a method for ensuring a tunable and stable non-unital evolution of the circuit was still lacking. We propose an algorithm for inducing damping by applying a controlled rotation to each qubit in the reservoir. It enables tunable, circuit-level amplitude amplification of the zero state, maintaining the system away from the maximally mixed state and preventing information loss caused by repeated mid-circuit measurements. The algorithm is inherently stable over time as it can, in principle, process arbitrarily long input sequences, well beyond the coherence time of individual qubits, by inducing an arbitrary damping on each qubit. Moreover, we show that quantum correlations between qubits provide an improvement in terms of memory retention, underscoring the potential utility of employing a quantum system as a computational reservoir. We demonstrate, through typical benchmarks for reservoir computing, that such an algorithm enables robust and scalable quantum random computing on fault-tolerant quantum hardware.