Hi,
Is there an impact on Accuracy in regard with the choice of the solver in a LogisticRegression ?
Hi,
Is there an impact on Accuracy in regard with the choice of the solver in a LogisticRegression ?
Since the loss function of LogisticRegression
is convex, until you get to full convergence, all the solvers will get to the same solution. So the question would be: why do you have so many solvers?
Basically, some solvers will be more efficient than other and may have been added later. A second possibility is that some solvers only work with some combination of parameters. In LogisticRegression
, it is usually linked to using sparse vs. dense data and the type of penalty L1/L2/ElasticNet.