The case for biologically plausible backpropagation

A researcher says their work convinced them that backpropagation—a method for training neural networks—is biologically plausible and can be implemented through vector-error feedback. They say their team hoped alternative learning rules would improve on backpropagation on ImageNet, but searches across thousands of rules did not deliver that improvement. The researcher also cites Francioni et al.’s 2025 Nature paper as confirmation of vector-error feedback.