This page consists of performance results of several MaTEx Algorithms. Each node consists of 16 cores We have used up to 4096 cores for performance evaluation. In several cases, we compare the performance with libsvm, which we have enhanced to use OpenMP for multi-core computing. We are planning to integrate this enhancement with MaTEx in the future release.
Support Vector Machine - SMO comparison with libsvmPerformance of MaTEx SMO algorithm with UCI Higgs Boson (2.6M/11M samples) dataset (4K cores)
Performance of MaTEx SMO algorithm with Offending URL dataset
Performance of MaTEx SMO algorithm with Forest Cover dataset
Performance of MaTEx SMO algorithm with MNIST dataset