Cache for ExampleSets?

Hi,

I have been wondering if there is any chance of caching the ExampleSets between multiple runs. In my case, the loading of the sparse data files takes lots of processing time every run but the data files do not change. So some kind of caching would be great to speed things up? Has this already been discussed or is there another solution to avoid reloading sparse files every run beside sql?

Greetings,
Harald

Find more posts tagged with

AI Studio

Accepted answers

All comments

land

Hi Harald,
did you try to save it into the repository? Might speed things up a lot...
Caching is in fact an issue, but this is not planned for the client version of RapidMiner.

Greetings,
Sebastian

harri678

I made a little benchmark and the "Read AML" of a sparse file is faster than store/retrieve repository.
sparse-file-specs: 7200 examples, 155340 attributes (16Mb .dat, 11Mb .aml, approx. 90% sparse)

I use "Read AML" and "Store" to save the data into the repository and made several loading-only tests to eliminate caching. These are the results:


          Retrieve Repo     Read AML (sparse)
1. run:   02:10             00:18
2. run:   02:03             00:19