**Notebook pédagogique pour comprendre le machine learning**

```
import numpy as np
import matplotlib.pyplot as plt
```

```
%matplotlib inline
```

# What is machine learning ?

### When we need Machine Learning

When the business logic is too complicated to be directly coded like for voice recognition, or logic is hidden like for customer behaviour on website, machine learning can be used to find this mysterious business logic.

## How it “learn” ?

First we select a mathematics model parametrable like a polynom or a decision tree. Then an optimizer will compare output given by model and output expected, it will tun parameters to minimize this error.

## How many machine learning exist

Many! There are two kinds :

**classifier:**according to input, it will try to select an category as output. Ex: try to find if customer will buy someting, sort cat images, etc**regression:**according to input, it will try to predict values. Ex: all forecast purpose, …

# How to represent business to learn ?

Represent a whole dataset to be visualize quickly, is complicated with real user case. Can you quick see a classifier dataset when you look a huge table full of numbers ?

But if you chose machine learning such as input is a position and output is a color. Then each prediction becomes a pixel and the whole dataset an image.