Extreme Memorization via Scale of Initialization
This work addresses the problem of understanding and controlling memorization vs. generalization in neural networks for machine learning practitioners, though it is incremental in exploring initialization effects.
The researchers investigated how varying initialization scales in SGD affects implicit regularization, leading to either good generalization or extreme memorization of training data, with specific activations like sin showing pronounced effects and ReLU's behavior linked to loss functions. They found that larger initialization scales cause misalignment of gradients and representations within classes, and developed an alignment measure that correlates with generalization in deep image classification models.
We construct an experimental setup in which changing the scale of initialization strongly impacts the implicit regularization induced by SGD, interpolating from good generalization performance to completely memorizing the training set while making little progress on the test set. Moreover, we find that the extent and manner in which generalization ability is affected depends on the activation and loss function used, with $\sin$ activation demonstrating extreme memorization. In the case of the homogeneous ReLU activation, we show that this behavior can be attributed to the loss function. Our empirical investigation reveals that increasing the scale of initialization correlates with misalignment of representations and gradients across examples in the same class. This insight allows us to devise an alignment measure over gradients and representations which can capture this phenomenon. We demonstrate that our alignment measure correlates with generalization of deep models trained on image classification tasks.