Machine Learning is the Science of getting computers to act without being explicitly programmed
- Andrew NG
Introduction
Machine learning, as the term suggests, pertains to the realm of machines, when I first encountered the phrase I thought it involves mechanical machines to some extent but that's not exactly the case with Machine Learning. Simply put, it is the art of enabling computers to learn without being explicitly programmed. In this blog, we'll be discussing Machine learning in a beginner-friendly approach.
What is Machine learning?
Before delving into understanding Machine learning, Let's explore how exactly humans learn. Take the example of a newborn: at birth, it lacks knowledge of how trees or humans or animals look like when it was born. However, as days pass it encounters various instances of trees, humans and animals. Over time it starts noticing trees has green leaves and brown branches, humans possess two eyes and two hands and animals typically feature a tail and four legs. By encountering a diverse range of trees, humans and animals, It learns to differentiate between these entities.
Now, let's connect this to the concept of machine learning. In machine learning, a similar approach is adopted. A large volume of data is fed into a machine-learning model. This data serves as the "experiences" for the model to learn from. By analyzing and processing this data, the model gains knowledge and develops an understanding of patterns and relationships within the data. Subsequently, when presented with new data or situations, the model utilizes its acquired knowledge to make predictions or draw conclusions about future outcomes.
Essentially, Machine learning refers to training a model with a vast amount of data and letting the model learn from the data and make predictions for future outcomes.
How does it work?
In traditional programming, we write explicit instructions for performing a given task. However, this differs from the state-of-the-art technology in machine learning. In machine learning, we allow the computer to learn from the provided data and take actions based on that information. The image below will further elucidate this concept.
How computer learns?
To facilitate computer learning, specific mathematical algorithms are employed. We shall delve into these algorithms in our forthcoming blog posts.
Types of Machine Learning
Like every single thing in the universe, Machine learning is also classified into three types. They are
Supervised Learning
Unsupervised Learning
Reinforcement Learning
Supervised Learning
Supervised learning is much like helping a child learn. We guide it by giving clear examples, just as we feed a baby with a spoon. The model learns from these examples and builds its understanding based on them.
A common example of supervised learning is how computers sort emails into spam or not spam. Initially, the model is given many emails, each marked as spam or not spam. It learns from this labeled data and creates a model. When a new email arrives, the model uses what it learned to decide if the email is spam or not.
Supervised learning has many uses, such as recognizing text and speech, or even detecting brain tumors in medical images. It's like teaching a machine to make sense of the world, step by step.
Supervised learning is classified into two types
Regression
Classification
In my upcoming blogs, we will delve into the details of these types. To put it simply, regression helps predict continuous values like numbers with decimals or integers. On the other hand, classification is about sorting things into categories like 0 or 1, where the values are discrete and separate.
Unsupervised Learning
Imagine giving your computer a collection of pictures featuring cats and dogs, all without any specific labels. Remarkably, the computer distinguishes these images and classifies them into two distinct groups, relying on patterns it discovers. This is accomplished through unsupervised learning, which sorts these images into categories such as "cat" and "dog."
This accomplishment has numerous practical applications. One noteworthy example involves the organization of images within digital galleries. The computer's aptitude for recognizing similarities between images leads to an orderly and visually appealing gallery presentation.
Additionally, unsupervised learning lends itself to enhancing social media platforms. Algorithms employ this method to customize content according to individual user preferences, leading to a personalized and captivating feed.
In essence, unsupervised learning subtly elevates our digital interactions by streamlining image organization and enriching social media engagement.
Furthermore, unsupervised learning can be categorized into two main types: clustering and anomaly detection. In our upcoming blog posts, we will delve deeper into exploring these two intriguing concepts.
Reinforcement Learning
Consider the similarity between a reinforcement learning model and a young Asian child. Just as a child learns from consequences, this model absorbs insights through rewards and penalties, honing its decision-making abilities.
Much like the Asian child, who realizes that performing well on exams brings praise, and underperformance can lead to serious problems (though Asian parents may sometimes appear unimpressed even if the child scores well), the reinforcement learning model understands that certain actions lead to positive outcomes (rewards), while others lead to less favorable results (penalties).
This fundamental principle, rooted in learning through iterative trial and error, forms the core of reinforcement learning. Just as young learners adapt their strategies based on feedback, AI game players refine their tactics through rewarded choices, self-driving cars fine-tune their navigation through corrective signals, robots learn tasks by repeated experimentation, and recommendation systems adjust content based on user interactions, enhancing user engagement on platforms like Netflix, Instagram, and Twitter.
These instances underscore the pervasive role of reinforcement learning across various domains. Much like a child growing through experiences, these models navigate their environments, shaping their behaviors guided by the nuanced dance of rewards and consequences.
Summary
In a nutshell, Machine learning unlocks the power of computers to learn from data just as we learn from experiences. It's a journey that spans from guiding a model like a child in supervised learning to uncovering insights in unsupervised learning and teaching through rewards and penalties in reinforcement learning. In the upcoming blogs, we'll be discussing the math required for Machine learning and essential Python libraries as well as the algorithms.