A0571
Title: Helmholtz machine with differential privacy
Authors: Junying Hu - Northwest University (China) [presenting]
Abstract: Helmholtz machine(HM) is the classic hierarchical probabilistic model for building the probability distribution of perception data, and the wake-sleep(WS) algorithm has been widely used as a training algorithm. To prevent the attacker from restoring the training set data by using the trained HM model, a Gaussian mechanism is introduced to the WS algorithm to propose a wake-sleep algorithm based on differential privacy (DP-WS) and use DPWS to train HM to get the HM model with privacy protection, named DP-HM. Rigorous proof of the privacy guarantee is provided. In addition, the experiments on MNIST and Bio-ID face datasets show that the DP-HM model can be trained under a modest privacy budget and still have acceptable model quality.
