Representational Drift in Neural Networks: What Backpropagation and Hebbian Learning Reveal
A study by an independent researcher found that neural networks trained with backpropagation and a hybrid Hebbian rule develop divergent internal representations despite achieving similar accuracy, with representational …