Backprop Alternative: Augmented Lagrangian Predictive Coding

Automated excerpt

Compared to PC, PC-ALM introduces dual neurons (Lagrange multipliers) per layer, making each layer's local recurrence a PI feedback controller. Performance differences between PC and PC-ALM become apparent in deep-narrow networks. PC-ALM: a local learning method for training 1000-layer neural networks PC-ALM successfully trains 1000-layer MLPs on MNIST. PC-ALM consistently narrows the gap between standard PC and global backprop. Credit magnitude across layers during inference.

Selected automatically from source text; not independently written or fact-checked. Read the original for full context.