Does anyone know which java libraries (sites with java codes) for implementing-optimization algorithms (estimating MLE of a simple regression model) based on Nelder and Mead (1965), BFGS or other "good" optimization algorithms (a java version of R's OPTIM-function- discrete fourier transforms- fast fourier transformJTB

commons-math

Thanks for reply, quantmeh.I will look into this packages.

colt

QuoteOriginally posted by: axsaxscoltcolts not supported, it's a dead project. its code base was ported to commons-math. sometimes i still use colt as it's very thin. commons is somewhat heavy.

Linear algebra in Java is painful. It's neither convenient nor fast.

QuoteOriginally posted by: katastrofaLinear algebra in Java is painful. It's neither convenient nor fast.At OpenGamma we are aware of and are working on these problems trying to improve speed of java maths with the OGMaths project. The project is open source and part of our Platform. The platform is available at http://developers.opengamma.com/. The platform version 1.0 should be released on 2 April.We also plan to release a procedurally orientated high level API to try and make it easier to perform mathematical operations in Java and to aid translation from more mathematically orientated languages to java.

To be honest, I think a good linear algebra library in Java should give up on object orientation. Just pass arrays around, and call Lapack-style functions on them. Occasionally one could use a light-weight adaptor to index the matrix in a 2d way, but only when constructing it.And store your functions in a class called "LA", to make it short.

QuoteOriginally posted by: katastrofaTo be honest, I think a good linear algebra library in Java should give up on object orientation. why would you need OO in linear algebra?

QuoteOriginally posted by: quantmehQuoteOriginally posted by: katastrofaTo be honest, I think a good linear algebra library in Java should give up on object orientation. why would you need OO in linear algebra?One reason is that it's easier to pass the object representing a matrix around, instead of separately the data pointer, row size and column size (what if you confuse their order?).But the penalty Java makes you pay for this is so huge, that I'd prefer to send a nice plain double[] array and the size separately.