tag:blogger.com,1999:blog-7181711759016870742.post7237096827105790352..comments2022-05-10T12:26:08.070+02:00Comments on Thomas Jungblut's Blog: Nonlinear conjugate gradient method in JavaThomas Jungbluthttp://www.blogger.com/profile/07157841886768146088noreply@blogger.comBlogger9125tag:blogger.com,1999:blog-7181711759016870742.post-1506364901241290422014-12-14T00:26:19.655+01:002014-12-14T00:26:19.655+01:00I see, maybe that's where my code is making so...I see, maybe that's where my code is making some problems. I'll def. have to keep eye on that. The fmincg in itself is very similar, except that I use Mathnet Numerics library for matrix/vector operations and you use your own library. :) I'm still looking for anything I've accidentally changed in the fmincg, rest of the code is same as exercise 4 from machine learning course..but i don't exclude the possibility that I've made some other mistake there. I'll def.share if this works. [can't share if it's wonky :D ] Thank you so much for your help. :)Anonymoushttps://www.blogger.com/profile/04890656172152021115noreply@blogger.comtag:blogger.com,1999:blog-7181711759016870742.post-60682772720446725342014-12-12T21:36:59.329+01:002014-12-12T21:36:59.329+01:00final is only there to not accidentally change the...final is only there to not accidentally change the reference (because there are multiple of the same kind, so easy to screw up). Similar to const in c#. I'd be eager to see the C# implementation (if you want to share obviously).Thomas Jungbluthttps://www.blogger.com/profile/07157841886768146088noreply@blogger.comtag:blogger.com,1999:blog-7181711759016870742.post-38466971692223018232014-12-12T21:33:56.409+01:002014-12-12T21:33:56.409+01:00Hi Thomas,
I've stumbled on your blog while l...Hi Thomas, <br />I've stumbled on your blog while looking for some sort of implementation for Fmincg function, since matlab code looked all confusing to me. I'm using your code, just adapted for C# . Thanks for sharing it :)<br /><br />I'm just a bit confused about keyword 'final' for your tuples. My implementation is a bit different, and I don't use tuple of any kind, but I wonder why is tuple variable final ? I tried understanding it, but I'm not familiar with Java enough to understand this. Actually, I don't see why is that variable final, is it about Java itself or you purposely wanted to make that tuple constant? <br /><br />I'm sorry if the question is silly one. I hope you see this. My fmincg doesn't lower J cost function by much even tho it should on say 100 epochs. <br /><br />All help appreciated. :)Anonymoushttps://www.blogger.com/profile/04890656172152021115noreply@blogger.comtag:blogger.com,1999:blog-7181711759016870742.post-67083680825138694382013-09-18T19:58:26.235+02:002013-09-18T19:58:26.235+02:00Thomas Jungblut, I have some problem with your cod...Thomas Jungblut, I have some problem with your code. Can you send me work java progect with f(x) = (4-x)^2+10 on mail?<br />My mail: krekota@yandex.ruДимаhttps://www.blogger.com/profile/10844322636315997173noreply@blogger.comtag:blogger.com,1999:blog-7181711759016870742.post-42001981300322616042013-08-08T05:43:37.416+02:002013-08-08T05:43:37.416+02:00Hey Andy,
I used my own math library (https://git...Hey Andy,<br /><br />I used my own math library (https://github.com/thomasjungblut/tjungblut-math). FminCG uses only the "de.jungblut.math.DoubleVector" interface- so you can implement it with practically every existing math library you have at hand. Thomas Jungbluthttps://www.blogger.com/profile/07157841886768146088noreply@blogger.comtag:blogger.com,1999:blog-7181711759016870742.post-48367374175689787302013-08-08T00:37:31.742+02:002013-08-08T00:37:31.742+02:00Very Nice.
I have some octave code I need to port...Very Nice.<br /><br />I have some octave code I need to port to Java and need fmincg like functionality. I'll check out your version. just out of curiosity what package are you using for general linear algebra support? I need basic stuff like multiply, transpose, member wise operations like octave .*, .+<br />random initialization, ...<br /><br /><br />Anonymoushttps://www.blogger.com/profile/09125717892996696840noreply@blogger.comtag:blogger.com,1999:blog-7181711759016870742.post-48823329424734270692012-07-19T08:55:31.860+02:002012-07-19T08:55:31.860+02:00Interesting idea, however I think that you can do ...Interesting idea, however I think that you can do this much more efficiently with specialized libraries than I would do in plain c++. <br />Also I have no experience in compiling native code for octave, so this would be much more of a research project and not really usable.<br /><br />The code in Octave's optim module (http://octave.sourceforge.net/optim/function/cg_min.html) is also interpreted.Thomas Jungbluthttps://www.blogger.com/profile/07157841886768146088noreply@blogger.comtag:blogger.com,1999:blog-7181711759016870742.post-26403624502090138382012-07-19T00:04:50.316+02:002012-07-19T00:04:50.316+02:00Have you considered rewriting the fmincg.m functio...Have you considered rewriting the fmincg.m function in C++ so that it could be used as a compiled .oct function in Octave. I, for one, would be really interested in this.Dekaloghttps://www.blogger.com/profile/07016889838486083128noreply@blogger.comtag:blogger.com,1999:blog-7181711759016870742.post-42176595512840708322012-07-18T18:05:42.471+02:002012-07-18T18:05:42.471+02:00Good post that I think our core audience of Java d...Good post that I think our core audience of Java developers would really like to read. Would you be interested in having this featured in Javalobby at DZone.com? If so, contact me at egenesky@dzone.comDZONEMVBhttps://www.blogger.com/profile/02587941708923838103noreply@blogger.com