Closed guidopetri closed 4 years ago
We're downsampling after vectorizing, but computationally, it makes more sense to downsample before vectorizing. This way we get fewer features and thus less of a memory load when training models.
We're downsampling after vectorizing, but computationally, it makes more sense to downsample before vectorizing. This way we get fewer features and thus less of a memory load when training models.