I made this PyMC3 model that performs fairly well for my problem. I wanted to test it against more traditionnal algorithms. It performs approximately as good as an un-tuned xgboost for my problem in terms of absolute error and R2, outperforms Lasso and Ridge models. What I really like about my PyMC3 model is the fact that I have the uncertainty of my predictions. But then I had the idea of fitting multiple models and averaging their predictions. Could I integrate the results of other algorithms in my PyMC3 model and adjust my confidence in consequence?