introduction_to_ml_with_python icon indicating copy to clipboard operation
introduction_to_ml_with_python copied to clipboard
verified publisher icon amueller

Issue with Input[39]

Open binodmainali opened this issue 5 years ago • 3 comments

The example provided in input[39] gives the following output Screenshot 2019-12-23 at 3 44 05 PM

Opposed to what is displayed in the book for LogisticRegression, I am new to the course and the boundary which separates class 0 and 1, the slope of the line seems to be inverted Output from the book Screenshot 2019-12-23 at 3 49 16 PM

binodmainali avatar Dec 23 '19 10:12 binodmainali

Thank you for the issue. Which version of sklearn are you using? Can you please give the output of sklearn.show_versions()?

amueller avatar Dec 23 '19 17:12 amueller

sklearn.show_versions()

System: python: 3.7.5 (default, Oct 25 2019, 10:52:18) [Clang 4.0.1 (tags/RELEASE_401/final)] executable: /Users/binod/.conda/envs/machine/bin/python machine: Darwin-18.7.0-x86_64-i386-64bit

Python dependencies: pip: 19.3.1 setuptools: 42.0.2.post20191203 sklearn: 0.22 numpy: 1.17.4 scipy: 1.4.0 Cython: None pandas: 0.25.3 matplotlib: 2.2.4 joblib: 0.14.1

Built with OpenMP: True

Python version 3.7.5

binodmainali avatar Dec 25 '19 12:12 binodmainali

Ok so the issue is that the solver in scikit-learn changed from liblinear to lbfgs. However, the new solver is more accurate. If the data was scaled, the two solvers would give similar results. I think the best thing to do would probably be to scale the data and make this less sensitive to the solver choices for now.

amueller avatar Dec 27 '19 18:12 amueller