interesting, would love to hear more about this. ive been struggling with noisy datasets in my own work recently. https://www.reddit.com/user/wolfunderdog45