title: “Machine Learning” excerpt_separator: “” categories:
- blog tags:
-
digital-humanities
- projects
Project title: Machine Learning
This was one of the greatest podcasts I’ve listened to about machine learning. The topic of machine learning has remained highly relevant in recent years because it appears in many different forms and applications.
One of the easiest ways to understand machine learning is to think about how a computer performs a task. In traditional programming, the algorithm gives the computer explicit instructions—step-by-step directions on how to complete a task. However, in machine learning, the goal is for the computer to learn how to do something on its own, based on past experience. Instead of being told exactly what to do, the computer figures out the task by learning from data.
There is a clear distinction between Machine Learning and Artificial Intelligence (AI). While AI is a broader field focused on making computers behave intelligently or rationally, machine learning is a subfield of AI that allows computers to learn how to behave intelligently by identifying patterns and learning from data. In recent years, computers have gained the ability to process huge amounts of data—something that wasn’t possible before—which has significantly advanced the field of machine learning. One of the most interesting ideas I learned from this podcast was reinforcement learning. I hadn’t thought about this before. Reinforcement learning is a process where computers learn from experience. For example, in a game like chess, you could simply program the computer with instructions on how to play. But if you allow it to learn by playing many games and losing, it will gradually figure out what moves lead to better outcomes and improve over time.
I was especially impressed to learn that reinforcement learning is inspired by the way humans train animals. Just like pets receive rewards for good behaviour and penalties for bad behaviour, reinforcement learning provides positive or negative feedback to help the computer learn the right actions. Smart cities are another example of how computer science is improving our world. Traditional traffic lights were controlled by timers. However, with the introduction of AI-powered traffic lights connected to cameras, the system can now detect when cars are approaching and at what times of day. This helps the system predict traffic patterns and adjust the lights accordingly, which improves traffic flow.
Handwriting recognition is another area that has seen great improvements. The computer is trained using labelled data—images of numbers, for example—so it can learn to recognize handwriting by comparing new inputs with its prior knowledge. A similar technique is used in email spam detection, where thousands of emails are labelled as either spam or legitimate. The computer learns the patterns and characteristics of each type and can then classify new emails accordingly. This type of learning is called supervised learning, where the system is trained using input-output pairs.