NOPROP:TRAINING NEURAL NETWORKS WITHOUT BACK-PROPAGATION OR FORWARD-PROPAGATION

NOPROP: TRAINING NEURAL NETWORKS WITHOUT BACK-PROPAGATION OR FORWARD-PROPAGATION

Qinyu Li Department of Statistics University of Oxford

Yee Whye Teh Department of Statistics

Razvan Pascanu Mila

http://arxiv.org/abs/2503.24322

一种借鉴flow match思想的无反向传播的方法，很有意思很新奇

结果：