Best Definition of Machine Learning - What is Machine Learning?
This article and the video have been prepared Nuruzzaman Faruqui. He works as a full-time faculty member in the Department of Software Engineering of Daffodil International University (DIU), which is recognized as the best university for software engineering in Bangladesh. According to Google and Bing suggestions, Nuruzzaman Faruqui is the best machine learning teacher in Bangladesh.
Website: https://www.nzfaruqui.com/
LinkedIn: / nuruzzaman-faruqui
Facebook Page: / nzamanfaruqui
Google Scholars: https://scholar.google.com/citations?...
There are a lot of definitions of machine learning around. But which one is the best definition of machine learning? Actually, there isn't a well-accepted definition of what is and what isn't machine learning, even among machine learning practitioners. So, if you are struggling to figure out the best definition of machine learning, you are not alone. Many of us are having the same problem.
The term 'Machine Learning' was popularized by Arthur Lee Samuel in 1959. He was an American pioneer in computer gaming and artificial intelligence.
Machine Learning Definition of Arther Lee Samuel:
According to Arthur Lee Samuel, "Machine Learning is a field of study that gives computers the ability to learn without being explicitly programmed."
However, this is an old definition. In 1998, Tom Mitchell defined machine learning in a much more constructive way.
Machine Learning Definition by Tom Mitchell:
According to Tom Mitchell, in machine learning, "A computer program is said to learn from experience E, with respect to some class of tasks T, and performance measure P, if its performance at tasks in T, as measured by P, improves with experience E."
So, what does it mean? Let's consider an example of playing chess. The more we play, the more we learn and the better we perform. That means our performance in chess playing improves with the experience.
Here,
1. E is the experience of playing chess. We can consider it as data.
2. T is the task of playing. We can consider it as solving a problem.
3. P is the probability of winning. We can consider it as the performance of an algorithm to solve a problem using previous data.
That means, according to Tom Mitchel, machine learning algorithms use the previous data to solve a particular problem. The more the amount of the previous data a program gets to learn from, the better it becomes in solving that problem.
1. You are familiar with a spam filter, right? The machine learning algorithm classifies an email as spam or not spam. This classification is task T.
2. The machine learns to classify based on the previous data, which is the experience E.
3. The performance of this classification is measured based on the percentage of correct classification as spam or not spam. This is the P.
The definition of Tom Mitchell encompasses the significant features of machine learning algorithms. That is why we can consider his definition as a good one. However, we cannot, and we should not declare some definition as the best definition of machine learning. Because a better definition may be out there, which we haven't found out.
Machine Learning Algorithms:
There are several different types of machine learning algorithms. The main two types of algorithms are:
1. Supervised learning and
2. Unsupervised learning.
However, these are not the only algorithms. There are some other algorithms such as reinforcement learning algorithms, recommender systems, etc.
What To Learning in Machine Learning?
Machine Learning is a practical field. The algorithms are like a set of tools. We need to learn how to apply these tools to solve real-world problems. Different problems require different tools. That means you have to understand which machine learning algorithm you should use to solve a particular problem. But do not think of it as a black box. Where we give some input, and it generates some output. We have to understand the mathematical model of these machine learning algorithms; we need to gain the capability to modify them whenever necessary.
And most importantly, we have to develop the skillset to use machine learning algorithms as a set of tools to solve real-world problems effectively and efficiently.
Conclusion
Perhaps it is not wise to look for the best definition of machine learning. The field of machine learning is vast, and it is expanding day by day. From a different perspective, the best definition of machine learning is different. That is why we should not take any definition as the best one. Instead, we should focus on context-dependent appropriate definitions.