Hello. Le mar. 18 juil. 2023 à 17:30, Dimitrios Efthymiou <efthymiou.dimitri...@gmail.com> a écrit : > > Hello everyone. I am working on the modularisation of > the legacy ml.clustering package to a new module: > commons-math-clustering. Some clustering classes > depend on stat.moment.Variance
In the new modules, there must be no dependency towards classes in the "legacy" module. Hopefully, [Statistics] will soon (?) contain a brand-new "Variance"[1] which the new module can depend on. In the meantime, there are maybe other issues that can be tackled.[2] > and some of > the ml.distance classes. I guess that "Variance" and "Distance" are not used for the same purpose. > 1--those distances belong to geometry probably and > not machine learning. Manhattan distance, for example. For the foreseeable future, [Geometry] will only deal with 1D, 2D, 3D. I.e. physical space with a low and fixed dimension. In machine learning, the space is routinely high-dimensional and the dimension varies from problem to problem. This must be handled at runtime by the implementation(s). > 2--should I move the distance package to the new > clustering module so that they are together or create a new > commons-math-distance module? It depends on whether the distance will be useful for more than just the clustering functionality. > Or put the distance classes > in the commons-math-geometry project? No, for the reason given above. Regards, Gilles [1] https://issues.apache.org/jira/browse/STATISTICS-71 [2] https://issues.apache.org/jira/issues/?jql=project%20%3D%20MATH%20AND%20status%20in%20(Open%2C%20%22In%20Progress%22%2C%20Reopened)%20AND%20text%20~%20%22cluster%22 --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@commons.apache.org For additional commands, e-mail: dev-h...@commons.apache.org
