Introduction to models — even toddlers could get it

When we hear the word ‘model’, our minds might go to something like

He’s gorgeous, and I love and appreciate models, but in data science, the concept is a little, or a lot different.

So, what are models?

When we call someone our model, we mean they are exemplary, and worthy to be copied. ML models are somewhat like that. Stay with me.

Behold, this piece of…. wait, what on earth is this?

You probably have never seen this before, but this is a shoe — the mojito shoe-and I’m not bluffing. (I can almost hear you laugh hysterically).

In case you didn’t see it clearly, this is the same shoe in different colors

Okay so no color can make this shoe look less weird, 🤣 but that’s not the point.

You might also be seeing this glorious…thing for the first time too- this is the infamous, resplendent banana shoe. (What is wrong with the world?)

The question is this:

Is this shoe mojito or banana shoe?

You most likely got the answer without breaking a sweat, and this is because I showed you the mojito shoe more than once, and the banana shoe once.

That’s modelling. A model is a file that you train to recognize different patterns. Models in machine learning are like your mind. It would have been easier if I asked you to spot a cat or a dog because over the years, your mind has been trained and you can boldly tell the difference.

The different techniques or approaches (known as algorithms) used in machine learning are:

1. Supervised learning
2. Unsupervised learning
3. Reinforcement learning

Supervised learning models

Here’s how you keep it in your memory — In supervised machine learning, you, or the person who collected the data is the supervisor, in the sense that you label the data. It’s just like how I showed you the mojito shoe, and then told you the name.

In this type of learning, the data is already tagged with the correct answer.

Types of supervised learning techniques include classification, regression.

Unsupervised learning

In unsupervised machine learning, you don’t label the data, the machine is smart enough to figure out hidden patterns in the data.

Reinforcement learning

You have a cat, Pearl. You want to teach her to walk. When you say ‘Pearl, walk ’ and she actually walks, you give her fish. If she doesn’t walk, you don’t give her anything until she responds appropriately.

That’s reinforcement learning. Reinforcement learning is concerned with how a software agent (your cat) takes actions (walks) in an environment (your house).

Don’t worry about the brevity — pause and think about what you’ve learnt today. (It’s a brief introduction and guide)

There’s more to come: what algorithms are and how they relate to models, what kind of algorithms you need to know, how to deploy models, and a plethora of ‘data-science-must-knows.’ Just follow me on medium and you’d see it each time I click the ‘publish’ button.

Oh, and BTW, I want to get someone the mojito shoe. Are you that someone?