pipette
ESEspañol

Quantum Reservoir Computing for Attention-like Selective Reproduction

Hadi Zadeh-Haghighi, Christoph Simon

Preprint

In the authors' words

Choosing which information to use as the context changes is a basic cognitive operation. We introduce a new task of this kind for quantum reservoir computing (QRC): a binary cue indicates which of two simultaneous input streams the reservoir must reproduce. In simulated spin reservoirs with a trained linear readout, infinitely many measurements (exact expectation values) let the reservoir reproduce the cued stream accurately. At a finite number of measurements \(N_M\), the score exceeds that of a linear readout of the raw inputs only at large \(N_M\) because of sampling noise. Performance increases with both entanglement and coherence in the computational basis, and then saturates. These results show that QRC can select and reproduce a cued stream, and they are a starting point for harder cognitive tasks.

Main resultLimitation the authors admit

Appeared: Thursday, October 15. arXiv. Preprint, not yet peer-reviewed.