We implement all machine learning algorithms, concepts and ideas here, with following principles:
-
Only use Numpy
-
All in modules/blocks
-
No dirty/garbage codes
-
As more matrix calculation as possible
-
As less for loops as possible
-
Write tests
-
Self-explained naming convention
Technically, we can create a mathematical model for any problems in the world with following elements:
-
Problem Definition
-
Historical Data
-
Model (Hypothesis)
-
Error (Loss, cost function, object function)
-
Optimizer