Boolean Function Learning

Boolean function learning asks how to recover a rule that maps binary inputs to binary outputs, and how well that rule will work on unseen cases. Approaches include Fourier and low-degree polynomial methods, threshold and circuit models, active or exact querying, and neural or symbolic rule learning. Researchers study guarantees in settings such as uniform, Gaussian, and correlated input distributions, including noisy labels and distribution shift. A central tension is that broad classes of rules may be statistically learnable but computationally difficult, while efficient or interpretable methods often rely on stronger assumptions about structure.