Closed simonmandlik closed 2 years ago
Currently, we're evaluating in ProductModel based on the ordering in samples. This will work as long as the same extractor is used. We should use model ordering instead to make sure we don't learn rubbish in m.W
ProductModel
m.W
Solved in https://github.com/CTUAvastLab/Mill.jl/commit/25bd4c90920d1634f367e479971d88870539b043
Currently, we're evaluating in
ProductModel
based on the ordering in samples. This will work as long as the same extractor is used. We should use model ordering instead to make sure we don't learn rubbish inm.W