Hi again Cesar,
Using the SMO learner, for MultiClass, with my large datasets, one question arised: how to map my RAM memory space to a value to use in the CacheSize from SMO algorithm?
I mean, how can I set the cache size to use 80% of my RAM?
In my computer with 4gb RAM, if I use the CacheSize = SampleSize, which in my case is 19K(with 130~ features), an exception is throwed for Insuficient Memory.
I think that if I can pass a size in kbytes instead of sample size to cachesize, would be easier and straight forward to use this amazing feature of SMO Algorithm.
Best regards,
Ben Hur