Open source Graphlab (Dato Core) toolkits does not provide supervised learning features

User 2124 | 8/4/2015, 5:38:48 AM

Hi,

I recently compiled and installed the open source version of Dato Core (Graphlab). However, in the graphlab.toolkits package, I didn't find any packages that support supervised machine learning (e.g., classifier, clustering, regression, ...). Also, the distributed.py file is also absent. The only thing contained in toolkits is the graphanalytics package, while in the commercial version, there are lot of machine learning packages. Does this mean Open source Graphlab does not support distributed execution and machine learning? Or the reason is simply I did not correctly configure Graphlab at compile time (e.g., MPI or HDFS, ...)?

Comments

User 1592 | 8/4/2015, 6:46:13 AM

Hi,

Then open source GraphLab Create engine includes all data engineering features for reading input data, cleaning it, making transformations, joining data from multiple sources and getting the data ready for the machine learning modeling. This is roughly equivalent to the functionality of Python Pandas but scaling to much larger datasets as we use out of core processing. The open source engine runs on a single multicore machine.

All the machine learning toolkits require a paid license. If you are a student you can apply for a free academic license. We provide an SDK so you are welcome to implement you own machine learning toolkits on top of the open source engine.


User 2124 | 8/7/2015, 5:16:22 AM

Hi Danny, Thank you for your clear answer. We are in a company and we must now consider purchasing a license.