I’ve been using PySpark’s Alternating Least Squares (ALS) algorithm for a recommendation system, and I’ve set the seed for initializing the latent factors to ensure reproducibility. However, I still observe different latent factors and predictions across multiple runs, even with the same seed set.