The PSL software from the University of Maryland and the University of California Santa Cruz
298
stars
100
forks
source link
Use first-order optimality conditions to measure convergnece of stochastic gradient descent reasoning. #327
Closed
dickensc closed 2 years ago
Use first-order optimality conditions to measure convergence of stochastic gradient descent reasoning.
This should make SGDReasoner tests more consistent as the change in the objective is not always an honest measurement of optimality.