Rami Mochaourab: Robust explanations for private support vector machines

Robust Explanations for Private Support Vector Machines. – We study the explainability of private support vector machines (SVM), where the privacy mechanism that publicly releases the classifier guarantees differential privacy. This notion of privacy relies on introducing uncertainty about the data through randomization. While privacy preservation is essential when considering sensitive data sets, there is a consequent degradation in the classifier accuracy. As a result, the explainability of predictions from such classifiers needs to be robust against the uncertainty introduced through the privacy mechanism.

Speaker Rami Mochaourab, RISE/Uppsala university