Variational Neural Annealing (VCA)

Citation: Hibat-Allah, M., Inack, E. M., Wiersema, R., Melko, R. G., & Carrasquilla, J. (2021). Variational neural annealing. Nature Machine Intelligence, 3(11), 952-961.

Overview

VCA is a variational framework that applies neural sampling techniques—particularly autoregressive models—to approximate Boltzmann-like distributions and perform combinatorial optimization via annealing dynamics.

Strengths

• Autoregressive Sampling for Exploration: The use of autoregressive neural networks allows VCA to flexibly explore the solution space and avoid common traps of static distributions.

Weaknesses

• Inductive Bias in Target Distribution: By focusing on approximating the Boltzmann distribution, VCA imposes a strong modeling bias that may not align with optimal or problem-specific objectives.