Fast Private Parameter Learning and Inference for Sum-Product Networks
This work addresses privacy-preserving parameter learning for sum-product networks in distributed settings, representing an incremental improvement by applying secret sharing instead of encryption for efficiency.
The paper tackles the problem of learning sum node weights for sum-product networks with horizontally partitioned data while preserving participant privacy, achieving this through secret sharing and a novel integer division method for approximate real divisions, and also demonstrates simple and private inference using the learned model.
A sum-product network (SPN) is a graphical model that allows several types of inferences to be drawn efficiently. There are two types of learning for SPNs: Learning the architecture of the model, and learning the parameters. In this paper, we tackle the second problem: We show how to learn the weights for the sum nodes, assuming the architecture is fixed, and the data is horizontally partitioned between multiple parties. The computations will preserve the privacy of each participant. Furthermore, we will use secret sharing instead of (homomorphic) encryption, which allows fast computations and requires little computational resources. To this end, we use a novel integer division to compute approximate real divisions. We also show how simple and private inferences can be performed using the learned SPN.