Linear and Logistic Regression in 60 lines of Python - Machine Learning From Scratch 04

Key Takeaways

hi everybody welcome to a new Python tutorial so in this video I will simply refactor the code of the last two videos so if you haven't watched the previous two videos about linear regression and logistic regression then please do so so now if you compare the code then you will see that it's almost similar so both classes have exactly the same init method and both have almost the same fit method so in both classes we init our parameters and then we do the same gradient descent except that in the linear regression we have simply this linear model for our approximated Y and in logistic regression we also have this linear model but then we also apply the sigmoid function and this same difference is in the predict method so linear regression we simply apply the linear model and in logistic regression we apply the linear model and then the sigmoid function and then say if it's 1 or 0 and we have this helper function but a lot of this code is similar so let's refactor this and yet let's create a base class and call it base regression and our two class classes will be derived from this base regression so let's say linear regression derived from base regression and the same for our logistic regression and then we can cut the init method and put it in the base regression because this is the same for both so let's cut this here and then we also cut the fit method so this is almost the same so we don't need this here and put it up here the only thing now that we have - that is different is the Y predictor because one time we need simply the linear model and one time we need the linear model and the sigmoid function so let's call let's create a helper function and call it underscore approximation which will get our data X our samples and then it also gets the weights and the bias and in our base class we will raise a not implemented error so this half has to be implemented in the derived classes so now in the linear regression class let's also create this and here we will implement this and in our linear regression model this is simply the linear model so we can return this return the dot product of X and W plus the bias and then we also have to or we can cut the fit method from our logistic regression method and implement the approximation method so here let's copy this so so let's remember that we need to have the linear model and then apply the sigmoid function so let's cut this and let's create our underscore approximation nest method with self XW and the bias and then we create our linear model which is numpy dot x w+ bias and then we apply the sigmoid function and return it so let's return self dot underscore sigmoid of our linear model so this is the approximation and now the predict method is a little bit different so let's in our pace class let's define this predict method and here we will implement a helper function and we call it underscore predict which will get self X and also W and B and also in the base class here we will simply raise a not implemented error so this now has to be implemented in the derived classes so we call and return this in our predict method in the base class so let's return self dot underscore predict and with test samples and then now with our calculated weights and the calculated bias and now in our linear regression class we define this underscore predict which will get X W and bias so and here it will get so this will be the same code so the dot product dot product of X and W plus the bias and we can return this in this line so we don't need this and then also in the logistic regression we define this is underscore predict with XW and B and then we have the same code here so we use W and B then apply the Sigma each define if it's 1 or 0 and return it and now we can copy this here and put it in the same file here and then we are done so now we have two models the linear regression and the logistic regression just in 60 lines of Python and it looks much cleaner now so yeah that's it I hope you enjoyed this tutorial and see you next time bye