Symbolic Knowledge Extraction from Opaque Machine Learning Predictors: GridREx & PEDRO