1 paper · 1 filter
Yongxin Chen, Tryphon Georgiou, Michele Pavon
We show that the continuous-time gradient descent in Rn can be viewed as an optimal controlled evolution for a suitable action functional; a similar result holds for stochastic gra…