You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@systemml.apache.org by "Manoj Kumar (JIRA)" <ji...@apache.org> on 2016/06/03 19:06:59 UTC
[jira] [Updated] (SYSTEMML-751) Support functions as arguments
[ https://issues.apache.org/jira/browse/SYSTEMML-751?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Manoj Kumar updated SYSTEMML-751:
---------------------------------
Description:
Hi,
How difficult is it to support functions as arguments? (If not done already)
One real use-case would be to just implement a modular lbfgs, newton-cg etc as in (http://docs.scipy.org/doc/scipy/reference/generated/scipy.optimize.fmin.html#scipy.optimize.fmin)
I believe this lessens burden on code maintenance and brings in new users, because every time I implement a new algorithm, I just need to check that my function (and gradient) implementation is correct.
I ultimately would like to do this.
source("lbfgs.dml") as lbfgs
f = loss_function() {
}
w_init = B = matrix(“0”, rows = 1, cols = 10)
w = lbfgs::optimize(w_init, f)
was:
Hi,
How difficult is it to support functions as arguments? (If not done already)
One real use-case would be to just implement a modular lbfgs, newton-cg etc as in (http://docs.scipy.org/doc/scipy/reference/generated/scipy.optimize.fmin.html#scipy.optimize.fmin)
I believe this lessens burden on code maintenance and brings in new users, because every time I implement a new algorithm, I just need to check that my function (and gradient) implementation is correct.
I ultimately would like to do this.
source("lbfgs.dml") as lbfgs
f = loss_function() {
}
w = lbfgs::optimize()
> Support functions as arguments
> ------------------------------
>
> Key: SYSTEMML-751
> URL: https://issues.apache.org/jira/browse/SYSTEMML-751
> Project: SystemML
> Issue Type: Wish
> Reporter: Manoj Kumar
>
> Hi,
> How difficult is it to support functions as arguments? (If not done already)
> One real use-case would be to just implement a modular lbfgs, newton-cg etc as in (http://docs.scipy.org/doc/scipy/reference/generated/scipy.optimize.fmin.html#scipy.optimize.fmin)
> I believe this lessens burden on code maintenance and brings in new users, because every time I implement a new algorithm, I just need to check that my function (and gradient) implementation is correct.
> I ultimately would like to do this.
> source("lbfgs.dml") as lbfgs
> f = loss_function() {
> }
> w_init = B = matrix(“0”, rows = 1, cols = 10)
> w = lbfgs::optimize(w_init, f)
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)