Methods that surpass ALS in a privacy-preserving advertising system using FHE

FHE Questions
1

I’m doing a research on a privacy-preserving advertising system, especially searching for ALS alternatives that will be compatible with FHE. I discovered some of them, but not competent enough to choose the best one of them and justify my choice. So I’m looking for someone who has enough knowledge to help me with this task. Here are the methods I discovered:

  • Factorization Machines (FMs; like LightFM, DeepFM);
  • Field-aware Factorization Machines (FFMs);
  • Graph Neural Networks (GNNs);
  • Polynomial Approximations of Neural Networks;
  • Homomorphic Encryption Optimized Neural Networks (HE-Friendly Networks);
  • Logistic Regression with Polynomial Kernels;
  • Tree-Based Models (Decision Trees, Random Forests, Gradient Boosting);