Is it possible to do incremental learning like scikit partial_fit in dato? required for large data sets of >10 million samples
User 940 | 10/13/2015, 5:21:45 PM
Unfortunately we do not support incremental learning right now. However, we strive to make all of our models very scalable such that they can support very large datasets.
Is there a particular model that is not scaling in the way you would like?
User 2356 | 10/14/2015, 5:58:32 AM
yes GL does not work for datasets of size >1 lac samples and also is not able to efficiently utilize multicore CPUs @piotr also check my other posts where I have explained each of the issues separately.