Quantum Reservoir Computing

Reservoir computing is a machine learning algorithm that takes a nonlinear dynamical system of sufficient complexity as a resourceful reservoir. The reservoir gains the capability of information processing by connecting to a trained model at its input or output, which could be simple linear regression or a more complicated neural network. One defining property of reservoir computing is the complex reservoir remains untouched in the processes of learning and computing. The role of the reservoir is to provide versatile nontrivial features. This means the reservoir computing naturally applies to the quantum simulation or noisy-intermediate-scale-quantum devices, which lacks universal quantum control to certain extent and is often imperfect. This has led to quantum reservoir computing. In our group, we are currently interested in addressing several questions including—(1) What is the designing principle for the quantum reservoir to best perform the computation power? (2) How to quantify the computation power of the quantum reservoir computing? (3) What would near-term quantum technology provide to real-world applications by quantum reservoir computing? Some progress has been made along these lines in our group.