# How to Learn Machine Learning from Scratch

Machine Learning was introduced by Arthur Samuel in 1959. Machine Learning is a form of systematic algorithmic machine learning that has been on the agenda of the world for years.

## How Can I Learn Machine Learning From Scratch?

There are many algorithms you can do to learn Machine Learning from scratch. To list them briefly below:

1) Problems; to produce intellectual solutions to solve it.
2) To be able to place mathematical functions in the background of the problem.
3) To synthesize various theorems we use in physics with mathematics.
4) To search the libraries used in the field of Supervised Learning and Unsupervised Learning.
5) Learning supervised learning and doing experiments for reinforcement learning.
6) To make reinforcements and mathematical exercises in the field of the coding language to be written (My coding language recommendation: Python),
7) To have theoretical knowledge about the working mechanism and mathematical formulas of algorithms.
8) You can learn all the above information ‘FREE’ from the internet…

There is indeed a lot of information and academic data on the internet in the field of Machine Learning. Entropy used in decision trees, gini; Many mathematical formulas, such as the Euclidean neighbor relation used in KNN, are available on the internet on the official description page of libraries or on mathematical websites. You will not need any university or course for this.

## Machine Learning and Mathematics

As you know, all models and algorithms used in Machine Learning are based on mathematics. As a matter of fact, Machine Learning is the autonomous form of mathematics taught to the machine. There are several advantages of using mathematics in this field. To briefly list them:

1) Compiling and collecting data and evaluating it within a certain scope.
2) Organizing the irregular data and putting them in algorithmic order.
3) Disciplined setting of provided actions and events in a single sequence and devoid of emotion (A purely logical sequence)

## Machine Learning and Models

True. To talk about the models that are important for Machine Learning, I would like to introduce you to the ‘Supervised Learning’ models that are most rooted in reason and logic.

Supervised Learning is a type of learning based on Machine Learning, which is known for containing the properties and tags of the data. Unlike Unsupervised Learning, it has labels. With this feature, it is known for having more paced and more cautious models.

The reason why Supervised Learning is easy to learn is because it is a type of function with a corresponding response. The formula we always use in mathematics is:

f(x) = ax+b+c…

The basic function shown above constitutes the basic logic of supervised learning, with F(x) = label, x = feature, according to mathematical rules. For this reason, the neuronal networks in the human brain will understand the completion of the bilateral interrelation more easily, and you will travel faster when starting machine learning from scratch.

## Machine Learning and Coding Language

Yes, dear tech readers. The coding language occupies a very large position in the field of machine learning. As a matter of fact, choosing a language with high accessibility and applicability will be the best option for you. The 3 popular Programming Languages ​​used in Machine Learning are as follows:

1) Python
2) C++
3) Java

Python, which is the most used language for machine learning among the programming languages ​​above, is a programming language that I recommend. As a matter of fact, the level of accessibility and applicability is quite high.

## Machine Learning and Algorithms

Algorithm.. Brings to mind the mathematical genius Carl Friedrich Gauss. Gauss is a scientist who made great contributions to the field of mathematics. He is a genius who developed a formula when his primary school teacher saw it as illogical to add numbers from 1 to 50 one by one. In fact, algorithms and formulas that he found years ago are used as a basic software element in systems containing high technology today. For example, the Radial Basis Function (RBF) is used in SVM models. The gamma parameter in it defines how far the effect of a single training sample reaches, with low values ​​meaning ‘far’ and high values ​​’close’.