1 paper
Yakir Oz, Gilad Yehudai, Gal Vardi +3
Current methods for reconstructing training data from trained classifiers are restricted to very small models, limited training set sizes, and low-resolution images. Such restricti…