Algorithms: The Brains of Machine Learning
Selecting the best algorithm for your ML problem is extremely important. This chapter will explore algorithms and meet the following objectives:
- PDF / 2,022,350 Bytes
- 49 Pages / 504 x 720 pts Page_size
- 9 Downloads / 249 Views
Algorithms: The Brains of Machine Learning Selecting the best algorithm for your ML problem is extremely important. This chapter will explore algorithms and meet the following objectives: •
Explain the terminology used by the scientists who create ML algorithms.
•
Show you how to select the best algorithm by considering multiple factors.
•
Summarize the three high-level styles of algorithms.
•
Provide a complete index list of CML algorithms so you can easily identify which style of ML a particular algorithm utilizes.
•
Present a decision flowchart and a functional flowchart that will help you to choose the best algorithm for your problem.
•
Present an overview of the seven most important ML algorithms.
•
Compare the performance of CML algorithms, including summary of CML and DL algorithms on the MNIST dataset.
•
Review the Java source code of popular algorithms.
4.1 Introduction When asked why they fail to deploy ML solutions, developers often express two main reasons: •
It takes too much manual work.
•
Algorithms and model creation are too complicated.
© Mark Wickham 2018 M. Wickham, Practical Java Machine Learning, https://doi.org/10.1007/978-1-4842-3951-3_4
177
Chapter 4
Algorithms: The Brains of Machine Learning
The manual work typically refers to data wrangling. In Chapter 2, I covered some tools that can help to make this simpler. Chapter 5 will present additional tools integrated as part of Java ML environments. With ML, it is hard to avoid manual work with data. I refer back to Mr. Silver’s interesting quote about expecting more from ourselves before we expect more from our data. In Figure 1-20, you saw that algorithms are rooted in the scientific domain. One of the main reasons developers shy away from deploying ML is that algorithm selection and model creation are too complicated. Fortunately, you can overcome the algorithm complexity issue by learning some basic principles and gaining an understanding of the scientific language associated with ML algorithms. This book and chapter will mainly cover CML algorithms. In Chapter 3, you accessed a DL algorithm via the Google Cloud Speech API because the cloud providers tend to focus their ML APIs on DL solutions.
ML-Gate 3 MLG3 is the phase during which you generate the ML model. The most important action during this phase is the selection and validation of the ML algorithm. This chapter will help you to choose the best algorithm and determine how it is performing. When you first embark on ML solutions, choosing the best algorithm seems somewhat arbitrary. In Chapter 5, you will see that it is actually quite easy to randomly select and apply any algorithm to a dataset. This is not a particularly good use of time. In Chapter 3, you saw disappointing results when the AWS Machine Learning wizard chose the wrong algorithm after misunderstanding the label data type. There is a conventional wisdom for algorithm selection. Answers to the following questions help to determine which algorithm is best suited for your model: •
How
Data Loading...
