Define the benchmark for spectral clustering 2.

spectra-cluster / spectra-cluster-2

A complete re-write of the spectra-cluster codebase

1 stars 2 forks source link

Open ypriverol opened 5 years ago

ypriverol commented 5 years ago

We have to design the spectral clustering benchmark, the following preconditions are needed for the test:

[x] A single machine, with local space for stable tests.
[ ] Different sizes of datasets.
[ ] Set of well-defined metrics for performance (time, cpu/memory consumption, ) and clustering quality.

ypriverol commented 5 years ago

I have to check my local Machine and we have:

Memory: 15.6GB Processor: Core i7 CPU 870 2.93GHz x 8 Disk: 500G

@jgriss let me know (with a +1) if you think this is fine and I tick the first precondition in the issue.

ypriverol commented 5 years ago

We should define datasets of the following size:

Performance datasets: 100k, 1M, 2M, 5M, 10M, 100M. Quality Datasets: Synthetic Peptide dataset.

jgriss commented 5 years ago

Suggestions:

ypriverol commented 5 years ago

Create a user for @jgriss in the machine for testing.