ConceptMutual Information Estimation
Mutual Information Estimation
A method for estimating the degree of statistical dependence between two random variables, commonly used in concept erasure to constrain the independence of representations from sensitive attributes. Exact computation is difficult for high-dimensional continuous variables, so neural estimators are required for approximation.
Timeline (last 90 days)
Oct 26
Nonlinear concept erasure methods (such as schemes based on adversarial training or mutual information estimation) erase more thoroughly in experiments, but they lack formal guarantees and often sacrifice downstream task accuracy.
Unverified50%