search

ai-se / SMOTE

0 stars 0 forks source link

using "hash"method to extract features would generate negative values #6

Closed WeiFoo closed 8 years ago

WeiFoo commented 8 years ago

Something wrong here.

azhe825 commented 8 years ago

Use this:

hasher = FeatureHasher(n_features=n_features, non_negative=non_negative)

this change will not affect classification result

timm commented 8 years ago

checking in here. is this running on "SE" data or "legal" data (i.e. specific data sets within stackexchange)

WeiFoo commented 8 years ago

this is SE data. and I need to check some details. right now, tfidf method works. so I use it to work on tuning stuff. I will go to check "hash" later on when all tuning methods running. This bug is not on my priority list now.

timm commented 8 years ago

@azhe825 is reporting that abs(hash) is not doing mich different to hash, anyway.

https://github.com/ai-se/e-disc/issues/54

so maybe we can close this thread?