Summary

In this chapter, we learned about applying classical machine learning algorithms in TensorFlow, without using neural networks. In the first section of the chapter, we learned about regression models. We explained how to train the models for linear regression with one or multiple features. We used TensorFlow to write the linear regression code. We also discussed that regularization is basically adding a penalty term so that the model does not overfit to the training data while learning the parameters in the training phase. We implemented Lasso, Ridge, and ElasticNet regularizations using TensorFlow. TensorFlow has some built-in regularization methods that we will study in the next chapters.

In the subsequent sections of this chapter, we learned about the classification problem in supervised machine learning. We discussed the model function, smoothing functions, and loss functions for binary class and multiclass classification. We used logistic regression in this chapter as that is...