Behaviour of the program with a set seed parameter in unexpected and needs to be aligned with requirements/common practice. There are two different unexpected behaviour patterns:
- Standard pycogaps current behaviour: all runs without a seed yield the same result, which inter alia numerically corresponds to
seed=0, which means the seed it set somewhere even if it's not provided by the user. Expected behaviour: runs without setting a seed differ. To reproduce: run pycogaps on modSim toy data without setting a seed and observe the resulting .obs and .var matrices are kept exactly equal. Run with seed:0 to obtain exactly same result. run with seed:42 to obtain a different result.
- Distributed pycogaps current behaviour: runs with the same seed yield different results. Expected behaviour: runs with the same seed produce exactly the same result. To reproduce: add
setParams(params, {'distributed':'genome-wide', 'seed':42}) and observe that results are different each time. Removing the seed from parameters does not affect the result randomness.
Behaviour of the program with a set
seedparameter in unexpected and needs to be aligned with requirements/common practice. There are two different unexpected behaviour patterns:seed=0, which means the seed it set somewhere even if it's not provided by the user. Expected behaviour: runs without setting a seed differ. To reproduce: run pycogaps onmodSimtoy data without setting a seed and observe the resulting.obsand.varmatrices are kept exactly equal. Run withseed:0to obtain exactly same result. run withseed:42to obtain a different result.setParams(params, {'distributed':'genome-wide', 'seed':42})and observe that results are different each time. Removing the seed from parameters does not affect the result randomness.