Measuring Forgetting of Memorized Training Examples
This work addresses privacy concerns in large-scale model training by revealing that early-seen examples may have enhanced privacy, though it is incremental in connecting memorization and forgetting.
The paper investigates the relationship between data memorization and forgetting in machine learning models, showing that standard models empirically forget training examples over time, which reduces susceptibility to privacy attacks on older data.
Machine learning models exhibit two seemingly contradictory phenomena: training data memorization, and various forms of forgetting. In memorization, models overfit specific training examples and become susceptible to privacy attacks. In forgetting, examples which appeared early in training are forgotten by the end. In this work, we connect these phenomena. We propose a technique to measure to what extent models "forget" the specifics of training examples, becoming less susceptible to privacy attacks on examples they have not seen recently. We show that, while non-convex models can memorize data forever in the worst-case, standard image, speech, and language models empirically do forget examples over time. We identify nondeterminism as a potential explanation, showing that deterministically trained models do not forget. Our results suggest that examples seen early when training with extremely large datasets - for instance those examples used to pre-train a model - may observe privacy benefits at the expense of examples seen later.