Guide to Machine Learning
Chapter 1 | Emergence
Artificial Intelligence (AI) has evolved from a research topic to a field with numerous applications. From weather prediction to cancer diagnosis, AI is transforming the world. Fun fact: The AI market was valued at over 60 billion dollars in 2020, and it is predicted to reach 250 billion dollars by the end of 2027. Since AI is booming, It is becoming extremely important to remain up to date with the techniques used in AI-based systems. At the core of present-day AI systems are Machine Learning (ML) models that can learn from data, recognize patterns and make decisions with minimal human intervention. This series is focused on learning all about ML.
What is Machine Learning?
According to scientific literature, Machine Learning is defined as:
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.
~ Tom M. Mitchell
Confused? Simply put, Machine Learning is the study of computer algorithms that improve over time as a result of experiences and the discovery of patterns in data.
The fundamental goal is for computers to learn on their own, without the need for human involvement, and to change their behaviour accordingly.
When a learning algorithm is combined with data, it produces a model. This model is then used to make predictions.
Exploring Machine Learning
The real-world problems that Machine Learning can solve are increasing day by day. To understand and solve a problem using ML, we need to know different ways in which it can be achieved and then choose the best option based on the problem at hand. Now that we’ve defined Machine Learning, let’s take a look at the first level of the ML universe.
In the coming days, we will be reading more about each of these types in-depth. In the meantime, I’ll explain when one of these options is used.
Classical Machine Learning: The simplest of algorithms that originated from pure statistics. It is used when the data is simple and the features are well-defined.
Example: Prediction of house prices based on features like square foot, number of rooms, furniture, whether a house has a garden or not, and so on.
Reinforcement Learning: It is a learning technique in which the agent interacts with its environment and learns from the consequences of its own actions. There is no initial learning dataset in this kind of technique.
Example: A reinforcement model which drives an opponent into checkmate in a chess game. It is only given basic rules of moving the pieces and an objective to defeat its opponent.
Ensemble Methods: It is a learning methodology that integrates multiple base models to build a unified optimal model for prediction. Used in most real-world challenges where quality is critical.
Example: Used in any situation where we need to make a decision that has some important consequence, we rely on the opinion of multiple experts to help make that decision. An abstract example would be the use of ML in the judicial system. Before deciding on a case, multiple models interact with each other to improve the decision and come up with the best possible decision.
Deep Learning and Neural Networks: Deep learning is a subfield of machine learning that uses an approach inspired by the human brain’s network of neurons to achieve machine learning. The goal of a neural network is to find patterns in data by employing nodes, or artificial neurons, as processing units. These nodes collectively form a layer and several such layers form a neural network. Typically used when the problem at hand is complex involving uncertain features and complicated data.
Example: Many fascinating technologies from self-driving cars to personal assistants use deep learning under the hood to make decisions.
We will be delving deep into these techniques in the coming chapters.
Too much learning, no fun?
Studying machine learning will help you create things that captivate others. This episode from “The Age of A.I.” is a must-watch.
