Types of Machine Learning Algorithms

Name: Types of Machine Learning Algorithms
Author: Briana Brownell

Explore the different types of machine-learning algorithms

Preview the first lesson free — get full access to all 4 lessons.

Course: On-Demand

Most of us encounter machine learning in some form every day, and those interactions can give us an understanding of how it works and when it doesn't. For example, many people are targeted with specific advertising after searching for something online, scrolling through video or TV show recommendations based on what they've watched, looking through their photos using keywords, sending an autocorrected text message, or phishing an email out of their spam folder that was incorrectly categorized. All these examples are based on machine learning, even though they might seem unrelated to each other. Algorithms are a series of instructions that solve problems, and mathematicians and computer scientists have created a wide variety of them to solve specific problems, such as the best move in tic-tac-toe, or getting you from an airport to your destination with the fewest stops. In the past, many algorithms that were developed were exact, meaning the developer needed to specify a set of rules for all possible cases. But these kinds of exact algorithms quickly run into problems. First, the problem that needs to be solved must be perfectly understood, which is fine for mathematical problems that can be fully described like tic-tac-toe. But what about when the system is only partially described? That's the case for many human-created systems. Consider the grammar of the English language. At over 1,700 pages, the definitive guide, a comprehensive grammar of the English language, considers itself the best and most complete description yet made of present-day English morphology and syntax, yet still notes the guide is comprehensive, not complete. Second, as the complexity of the system increases, so does the difficulty to solve it. Although the game of checkers has been played for 5,000 years, it was only in 2007 that checkers was considered solved. Checkers has 500 billion possible positions, and even with the world's fastest computers, the problem of solving checkers is extremely difficult because of the sheer number of cases that need to be considered. These difficulties made researchers think about the problem solving differently. Rather than using complicated methods and specifying a complicated set of rules, perhaps the problems could be solved by developing an algorithm that let the computer learn rules using data. For example, the machine could learn how to play checkers by playing the game and seeing what moves are associated with winning. Machine learning's reliance on data has meant its popularity has surged in the last decade. A few social and technological changes had made it easier than ever to create and deploy machine learning models and technology. Thanks to our increased use of technology, there's more data available to train machine learning models.

Intermediate Provider : Briana Brownell 4 Lessons · 8m in English

Course Description

Machine learning is one of the most exciting branches of data science. It can be used in a wide variety of contexts to solve many different problems, including classifying images, detecting spam, labelling images, segmenting customers, and making predictions. There are several types of machine learning, including supervised machine learning and unsupervised machine learning, and they can be used to solve very different kinds of problems.

In these lessons, you’ll learn about the history of machine learning and the difference between machine learning and other methods for problem solving. We will discuss the reasons why machine learning has been so successful in solving some of the most difficult problems like machine translation and image recognition. We’ll review the importance of data collection and common best practices in data collection including issues such as missing data, anomalous data, and falsified data. Finally, we’ll discuss the typical machine learning pipeline and the common stages that a business would go through in order to create and deploy a machine learning model.

What You'll Learn

Distinguish between supervised and unsupervised machine learning
Explain the historical impact of machine learning and why it has succeeded at hard problems like machine translation and image recognition
Describe the kinds of problems machine learning can be applied to, such as image classification, spam detection, customer segmentation, and prediction
Demonstrate the advantage of using machine learning compared to other problem-solving methods
Apply best practices in data collection, including handling missing, anomalous, and falsified data
Outline the typical machine learning pipeline and the common stages a business goes through to create and deploy a model

Key Takeaways

Machine learning is a branch of data science used in many contexts, including classifying images, detecting spam, labelling images, segmenting customers, and making predictions.
There are several types of machine learning, including supervised and unsupervised machine learning, which are suited to solving different kinds of problems.
Machine learning has been especially successful at solving difficult problems such as machine translation and image recognition.
Data collection is important, and common best practices address issues such as missing data, anomalous data, and falsified data.
Creating and deploying a machine learning model follows a typical pipeline with common stages a business goes through.

Frequently Asked Questions

What topics does this course cover?

The course covers an overview of machine learning, the types of machine learning (including supervised and unsupervised), data collection, and the machine learning pipeline. It also discusses the history of machine learning and the difference between machine learning and other problem-solving methods.

What is the difference between supervised and unsupervised machine learning in this course?

The course teaches you to distinguish between unsupervised and supervised machine learning, explaining that they can be used to solve very different kinds of problems.

What kinds of problems can machine learning solve according to this course?

The course describes problems such as classifying images, detecting spam, labelling images, segmenting customers, and making predictions, and notes machine learning's success with difficult problems like machine translation and image recognition.

Does the course address data quality?

Yes. It reviews the importance of data collection and common best practices, including issues such as missing data, anomalous data, and falsified data.

What skills will I gain from this course?

The course develops skills in machine learning, machine learning algorithms, machine learning methods, machine learning model training, automated machine learning, and decision tree learning.