Machine Learning Concepts
Get an overview of the types of machine learning and their applications.
In this lesson, we’ll review machine learning to help us understand the thought process behind this technology and how it's proving to be a game changer in every field.
What is machine learning?
Machine learning is a subfield of artificial intelligence (AI). It focuses on building algorithms and models that enable a computer to learn from data and make predictions for new data. In simple words, machine learning refers to teaching computers (machines) how to make decisions like a human without being explicitly programmed to do so.
The history of machine learning dates back to the 1940s and 1950s when researchers began exploring the idea of using computers to learn from data. In 1956 American psychologist Frank Rosenblatt proposed the first machine learning algorithm, Perceptron. At that time, development in machine learning technology was limited by computing power and data storage capabilities.
In the late 1990s and early 2000s, it gained traction with the advent of more advanced computer hardware and large amounts of data. New algorithms like decision trees, support vector machines, and neural networks have further increased the focus on this technology. This development led to breakthroughs in complex areas, such as chronic disease prediction and image and speech recognition, and it allowed for the widespread use of machine learning in industry and research.
Applications of machine learning
Today machine learning technology is easily accessible. It’s shaping our lives and making them easier in many ways. For example, have you ever wondered how a streaming service recommends movies and TV shows? Or how our smartphone camera can identify and tag people in our photos? These are just a few examples of how we use machine learning in everyday life.
In the healthcare industry, machine learning algorithms are used to diagnose chronic diseases, predict patient outcomes, and discover potential drugs. They’re also widely used in financial organizations for credit scoring, fraud detection, and algorithmic trading. Fast-moving consumer goods (FMCG) companies are using machine learning to forecast their demands and manage inventories. Machine learning is also very helpful in building autonomous vehicles, detecting traffic lanes for driverless cars, and optimizing delivery routes for shipments.
Machine learning technology continues to grow and evolve as computer hardware becomes more powerful and researchers develop faster and more advanced algorithms.
Types of machine learning
There are several types of machine learning. Based on the nature of the task, machine learning problems can be broadly classified into three main categories:
Supervised learning
Unsupervised learning
Reinforcement learning
Supervised learning
Supervised learning is the most common type of machine learning. It includes classification (e.g., classifying email into spam/not spam) and regression tasks (e.g., predicting house price from the number of bedrooms as input).
In supervised learning, we use a dataset that consists of input features (numerical, categorical, texts, or images) and a set of labels (response variable or target feature) to train a model that can make predictions on unlabeled/unseen data based on the patterns it learned from the training data.
Unsupervised learning
Sometimes instead of predicting from a dataset, we are interested in discovering common or hidden patterns, segments, or clusters in the data. This type of task comes under the category of unsupervised learning. In an unsupervised task, the model is not given explicit labels or targets. Instead, it must find patterns or relationships in the data on its own.
Unsupervised learning is very useful in exploratory data analysis (EDA), clustering, anomaly detection, and other applications where we don’t have a labeled dataset.
Reinforcement learning
Reinforcement learning (RL) is a type of machine learning where the goal is not to segregate data into clusters or make predictions but rather to teach machines or agents behaviors that maximize the reward signal and minimize the penalties. The agent receives feedback in the form of rewards or penalties and adjusts its behavior over time to maximize the cumulative reward.
RL is used extensively in the field of robotics (e.g., training robots to grasp an object), autonomous vehicles (e.g., self-driving cars navigating lanes), and video games (e.g., computers learning how to play chess).
In this course, we’ll only cover supervised and unsupervised learning in detail since the H2O package doesn’t have RL support.