What is Machine Learning?

What is Machine learning?

Machine learning is defined by Arthur Samuel as: “Field of study that gives computers the ability to learn without being explicitly programmed”.

How do computers learn?

Short answer: Using data (lots of it!).

Long answer: By using the patterns generated by complex algorithms after processing huge amounts of relevant data, to make predictions on a new set of data.

Alright. Now, I have heard other words like Artificial Intelligence and Deep learning. Are they the same as machine learning? Or are they different?

This is a popular visualization of AI, Machine learning and Deep learning

• AI, or Artificial Intelligence, is a broad field. It concentrates on enabling machines to perform human-like actions.
• Machine learning is an approach to AI which relies on training the machine on data to enable it to take its own decisions.
• Deep Learning is a subset of Machine learning, which uses neural networks (a network of algorithms arranged in a way such that it mimics the way the human brain works) to improve the result of a certain algorithm.

Types of Machine Learning (based on the type of data)

• Supervised Learning: In supervised learning, the data used for training has labels and features. eg: If the problem is to predict the price of a house based on its location and size, then the location and size are the features and the price is the label.
• Unsupervised Learning: The data has only features and no labels. eg: Clustering fruits into groups based on their colour and size.
• Reinforcement Learning: Reinforcement Learning algorithms interact with their environment (The problem the algorithm is trying to solve. In this case, we will assume that it is a maze) with a series of trial-and-error actions (like go straight, turn right etc.). Each action results in either a reward or an error (reward- closer to the end, error- farther from the end or an invalid action). In these scenarios, the algorithm’s main aim is to maximise reward. This eventually leads it to find the optimal pattern in learning, i.e. a way that helps it maximise reward.

Types of Machine Learning (based on the nature of the data)

• Structured data: The data is in the form of spreadsheet or data bases (rows and columns). eg: .xlsx files, .csv files and databases like SQL.
• Unstructured data: The data is comparatively unstructured in nature as there are no rows and columns. eg: Images, audio waves, Natural Language text (Human languages), Videos.

Types of Machine Learning Problems

• Classification: The algorithm predicts a given input to belong to one group or not. There are two types of Classification problems — Binary classification (Where there only two groups of data) and Multi-class classification (Where there are more than two groups in the data). This is a category of supervised learning.
• Regression: The model (algorithm) predicts a number for a given input. eg: Predicting the sale price of a computer based on its features. This too is a category of supervised learning.
• Clustering: This divides data into clusters. This is a category of unsupervised learning.
• Transfer Learning: The algorithm borrows the patterns another algorithm has found out from a similar dataset and uses it on our data.

Real-life examples of Machine Learning

Have you ever thought how YouTube just knows what you want to see? Or how Google auto-completes your searches? The answer to these questions is Machine learning.

Here are some day-to-day examples of Machine Learning:

• YouTube video recommendation
• Speech recognition and synthesis in virtual assistants like Alexa or Siri
• Natural Language translation
• Spam classification

These are just a few uses of ML and hope you are starting to see the power of ML.

So this is a brief overview of Machine Learning, its applications and its various types.

Alright, this is it for now.

Bye!