Machine Learning in 30 seconds
Introducing machine learning to non-practitioners is an interesting challenge. What seems to be intuitive for the initiated is often tough to grasp for the uninitiated. This article chronicles my experience of introducing AI to the larger public.
If you can’t explain it to a six year old, you don’t understand it yourself.
- Albert Einstein
A picture is worth a thousand words
Here are three cats:
There’s a jaguar, leopard and a cheetah in the picture. Let’s try to find the species of the cat.
Taking a good look at the cats and we notice the following:
Each of the cats have a different facial structure, texture and color to them. Moreover the differences are marked enough to distinguish the cats from each other.
The answer is:
Looking back at the features and noticing the correct answer, we’re now able to identify the cats.
Machine learning refers to the act of teaching computers to find patterns in data.
How do you teach computers to find patterns?
There are three main ways to find patterns:
- Supervised Learning. When an expert (usually human) explains the pattern to the computer. The explanation is often done through the use of “labels” and the process of generating the labels is often called as “annotation/labelling”.
- Unsupervised Learning. When computer finds the pattern by itself.
- Reinforcement Learning. When the pattern is learnt by the computer in a dynamic fashion by following a reward function.
