forum.alglib.net

Alglib is really developing and we are very happy with the current library. We are wondering if there are plans to add Bayesian algorithms and Markov Chained Monte Carlo algorithms at some point?

MCMC is often used for probit regression, which would be a nice addition to the logit model.

Also, the linear regression implementation seems to support a non-intercept form but the logit regression does not. Will it be possible to specify a non-intercept form for logit at some point?

Personally, I have always been interested in the data analysis. I think that future versions of ALGLIB will see more data mining algorithms (including Bayesian algorithms). However, priorities for the next six months are new languages (Python/IronPython versions will be released in December), interpolation/fitting, numerical optimization.

Some minor advances in data mining algorithms are possible, but significantly new functionality will be implemented only if some commercial user will pay for it. Right now our users (both commercial and not) are more interested in optimization than in data mining :) Maybe in the late 2011, when I'll have more people working on ALGLIB...

Non-intercept form of logit seems simple enough addition, but unfortunately, there are too many tasks to be solved in this year - and the end is near :)