Energy consumption of thermodynamic Bayesian logistic regression

Quantify the energy cost required by the proposed Bayesian logistic regression thermodynamic circuit to obtain N posterior samples, including precise scaling with parameter dimension d and batch size, and provide theoretically justified lower and upper bounds on work and heat dissipation.

Background

While the paper derives energy complexity for the Gaussian–Gaussian model, it does not provide an energy analysis for the Bayesian logistic regression circuit, noting that such a paper would be more involved and is outside the scope of the current work.

The conclusion explicitly identifies the quantification of energy consumption for the logistic regression protocol as an immediate extension and an open question, emphasizing its importance for fair comparison with digital methods and for assessing practical viability.