Book Image

Machine Learning Quick Reference

By : Rahul Kumar
Book Image

Machine Learning Quick Reference

By: Rahul Kumar

Overview of this book

Machine learning makes it possible to learn about the unknowns and gain hidden insights into your datasets by mastering many tools and techniques. This book guides you to do just that in a very compact manner. After giving a quick overview of what machine learning is all about, Machine Learning Quick Reference jumps right into its core algorithms and demonstrates how they can be applied to real-world scenarios. From model evaluation to optimizing their performance, this book will introduce you to the best practices in machine learning. Furthermore, you will also look at the more advanced aspects such as training neural networks and work with different kinds of data, such as text, time-series, and sequential data. Advanced methods and techniques such as causal inference, deep Gaussian processes, and more are also covered. By the end of this book, you will be able to train fast, accurate machine learning models at your fingertips, which you can easily use as a point of reference.
Table of Contents (18 chapters)
Title Page
Copyright and Credits
About Packt
Contributors
Preface
Index

Statistical modeling – the two cultures of Leo Breiman


Whenever we try to analyze data and finally make a prediction, there are two approaches that we consider, both of which were discovered by Leo Breiman, a Berkeley professor, in his paper titled Statistical Modeling: Two Cultures in 2001.

Any analysis needs data. An analysis can be as follows:

A vector of X (Features) undergoes a nature box, which translates into a response. A nature box tries to establish a relationship between X and Y. Typically, there are goals pertaining to this analysis, as follows:

  • Prediction: To predict the response with the future input features
  • Information: To find out and understand the association between the response and driving input variables

Breiman states that, when it comes to solving business problems, there are two distinct approaches:

  • The data modeling culture: In this kind of model, nature takes the shape of a stochastic model that estimates the necessary parameters. Linear regression, logistic regression, and the Cox model usually act under the nature box. This model talks about observing the pattern of the data and looks to design an approximation of what is being observed. Based on their experience, the scientist or a statistician would decide which model to be used. It is the case of a model coming before the problem and the data, the solutions from this model is more towards the model's architecture. Breiman says that over-reliance on this kind of approach doesn't help the statisticians cater to a diverse set of problems. When it comes to finding out solutions pertaining to earthquake prediction, rain prediction, and global warming causes, it doesn't give accurate results, since this approach doesn't focus on accuracy, and instead focuses on the two goals.
  • The algorithm modeling culture: In this approach, pre-designed algorithms are used to make a better approximation. Here, the algorithms use complex mathematics to reach out to the conclusion and acts inside the nature box. With better computing power and using these models, it's easy to replicate the driving factors as the model keeps on running until it learns and understands the pattern that drives the outcome. It enables us to address more complex problems, and emphasizes more on accuracy. With more data coming through, it can give a much better result than the data modeling culture.