Chicago Machine Learning Study Group Message Board › Mahout and k means methods
So I thought I would try this channel:
As I posted previously, I wondered if someone had managed to compile Mahout as a JAR? I have managed this with the Weka distribution, but the Maven build process for Mahout isn't as transparent.
I also wondered if you had found a best solution for the Twitter classification problem. I had read recently about an eigenvector approach, clustering, that reminded me of a Lagrangian-based approach I had seen on Video Lectures. I understand why an eigen vector approach would make sense, but wondered if it placed too many assumptions on the underlying process? On that note, I would point out that NIST has a project that recommends a graphical approach to help identify appropriate methodologies for analysis of this type of statistical data.