Thermodynamic cost of Quantum Extreme Learning Machines

Extreme Learning Machines (ELMs) are feed-forward neural networks with a single hidden layer and random couplings, trained via linear regression, saving time and resources. Their quantum counterparts (QELMs) use a quantum reservoir to build complex representations of classical data, with inexpensive, noise-robust performance.The Quantum Theory Group at UNIPA has demonstrated the advantages of QELMs for validating and certifying quantum systems without tomographic reconstruction. These algorithms apply to the optimisation of complex processes, quantum and classical, on NISQ machines, to the energetic characterisation of biochemical systems simulated on quantum computers, and to the optimisation of networks for communication and logistics.The research will analyse the thermodynamic cost of QELMs to establish their energetic sustainability, using quantum information thermodynamics, a field in which the group is a pioneer.