Performance analysis of an autonomous quantum thermal neuron

Performance analysis of an autonomous quantum thermal neuron

Seminar author:Loris Cros

Event date and time:06/26/2025 02:30:pm

Event location:Seminars Room, GIQ

Event contact:

Thermodynamic computing is a branch of analog computing that uses thermodynamic processes to perform computation. In previous work, a model was introduced that emulates the behavior of a perceptron using a network of qubits coupled to distinct thermal baths. It was shown that this system could reproduce any linearly separable binary function in its steady state. In this thesis, we present a modified design that replaces the qubit network with a single qudit. We demonstrate that this simplified system can still replicate the computational behavior of a perceptron. We analyze the system’s steady-state performance in terms of entropy production rate and classification accuracy, and we investigate its temporal evolution, comparing accuracy, speed, and entropy production. In the high accuracy domain, we see an exponential increase in precision as a function of the rate of entropy production.