Skin effect in quantum neural networks

A central assumption in statistical and condensed matter physics is that geometric boundaries and microscopic changes scarcely influence the global behavior of a physical system in the thermodynamic limit. For example, in Ginzburg-Landau's theory of continuous phase transitions, symmetries play a key role, while other features, such as boundary conditions, are deemed negligible. This paradigm has been severely challenged recently after the introduction of non-Hermitian (NH) topological systems and the discovery of the NH skin effect. The emergence of the skin effect in systems beyond lattice configurations and its potential impact in emerging areas of quantum technologies, such as in quantum neural networks (QNNs) and quantum computation, remain largely unexplored. Considering the well-known framework of quantum reservoir computing, we show how the performance of a QNN can exhibit a skin effect, even in the case of irregular networks.