Why is TIMi modeler so fast?
- Thread starter Artur
- Start date
We wrote from scratch all linear algebra routines to use the "CPU cache" better than in any other application. All our routines are vectorized and multi-threaded.The "Elastic Net" algorithm included inside TIMi is 100% written from scratch: no sub-standard code here! We have also our own proprietary in-memory compression routines to handle very large learning datasets (like in telecom and retail). for example, we can create extremely accurate models on a learning dataset that is a 2GB text file in a few minutes.