Machine Learning Explained in 5 Minutes

Hey guys , Jade here .

So you've probably heard the words machine learning a lot in the past few years .

And today I just wanted to talk to you about what it actually means .

So a lot of people think that artificial intelligence and machine learning are the same thing , but they're actually not artificial intelligence or A I , as the cool kids say is the entire field dedicated to making machines .

Smart machine learning is a division of A I .

And it's really cool because you don't actually have to program them machine to do what you want .

It improves or learns by repeating the same task over and over again , slightly tweaking the process each time until it's doing exactly what you want .

It's similar to how a person learns from experience .

So to give you an example , imagine your grandpa works at the local grocery store .

His job is to sort the apples from the bananas , but he has a bad back and complains all the time .

So you want to program a machine to do this job for him .

The first thing you do is to find some features it can use to distinguish apples from bananas .

One might be color , the redder , the fruit is the more likely it's an apple , the yellower .

It is , the more likely it's a banana .

This is a pretty good starting point .

But what if we encounter a situation like this ?

Obviously , color isn't enough .

We need another feature .

Another good one might be hardness , the harder the fruit , the more likely it's an apple , the softer , the more likely it's a banana .

We can choose a lot more features , but let's keep it at two to keep things simple .

Now that we've defined some features , we need to collect some data to train our machine .

So you go and get 50 apples and 50 bananas and label each one as item , one item two all the way up to 100 then you measure the color and hardness of each one note down whether it's an apple or a banana and feed your data into the machine .

Now it's time to test it out at first .

It kind of sucks .

Even though it's got its two features of color and hardness .

It doesn't know how much importance or as computer scientists say , wait to place on each of them .

The idea of weights in making a decision is probably the most central idea in machine learning .

And it's something we're all familiar with .

Say you're looking for a romantic partner and two things you're looking for are intelligence and a nice smile , but maybe you place more weight on intelligence than you do on a nice smile .

But also if they never smiled , that wouldn't be good either .

So you still place some weight on a nice smile .

Weights play a huge role in any decision .

That's why when your friend tells you , they found you a good match .

They're usually always wrong .

They may know what features you like , but they don't know the weights .

What separates machine learning from previous forms of artificial intelligence is its ability to figure out these weights .

It does this by comparing its results to the data that you gave it .

At first , your fruit sorting machine distributes the weight randomly .

It starts with item one , which happens to be a very unripe banana , which is of course , much harder than a regular banana .

If it places too much weight on hardness , it'll be like , OK , it's hard .

So it must be an apple .

But by comparing this to the data you gave it on item one , it'll realize it's made a mistake .

It readjusts so that less weight is placed on hardness and more weight is placed on color , but it overshoots .

And when it encounters a yellowish apple , it wrongly classifies it as a banana .

But again , by comparing its results to the data you gave it , it realizes its mistake and readjusts the weights .

Again , it repeats this process of comparing and readjusting until it's barely making any mistakes .

After some time , it's ready to take your grandpa's job at the local market .

This example was a simplified version of what Facebook is doing when it tags you in a photo .

Facebook learns your face by analyzing photos , which it knows are you by measuring the size of your eyes or the length of your nose .

And it uses this data to identify you in the next photo .

Actually , machines are getting pretty good at this in 2015 , a robot beat a human at image recognition for the first time .

And now they're developing robots that can clean your home for you .

How cool is that ?

If you'd like to know more about how this weight adjusting process works , make sure to subscribe to up and atom .

So you don't miss my next video on neural networks , which is really where all the magic happens .

So yeah , um until next time .

Have a good day .