How to avoid memory problem while processing huge table?

I have a huge observation table with around 30 Lacs of rows and 12 columns. While training knn classifier in 2016a version, I am getting errors related to memory. Is there any way to avoid this? I have tried to reduce rows but it's affecting the output quality.

Each row in table is a pixel and it's other values as features in columns. In one set of MRI scan, there are around 20 images of 512x512, I am loading one set for creating observation table. Is there another way to pass large amount of data to knn classifier?

Answer 1

[KSSV]

doc datastore, memmap, mapreduce.

[Nitinkumar Ambekar]

Thanks @Dr. Siva, one small query: Can I pass one of these to a function which takes `table` or `matrix`?