Introduction to machine learning

Training is a universal skill that is inherent in any living organism on the planet.

Learning is the acquisition of knowledge and skills through experience or study. This is what unites us and at the same time makes us unique. This is something that develops over time.


What if I say: “Machines can learn too”?

We live in an amazing time of development of robotization, artificial intelligence and machine learning. Machine learning is still a fairly new concept. We can teach machines how to learn, and some machines can learn on their own. Magic!

This article will introduce you to the basics of machine learning.
The most important thing you need to understand is that machines can predict the future based on the past.

What is machine learning?

Machine learning assumes that a computer recognizes patterns using examples, rather than being programmed using certain rules. These templates are contained in the data.

Machine learning is the creation of algorithms (a set of rules) that learn from complex functions (templates) from data and makes predictions.

It happens in3 stages :
1) Data analysis
2) Finding patterns
3) Prediction based on a pattern

Applying Machine Learning

An overview of where machine learning can be used.

  • Health care: predicting patient diagnoses for doctors
  • Social Network: Dating Site Compatibility Prediction
  • Finance: Credit Card Fraud Forecasting
  • E-commerce: customer churn prediction
  • Biology: search for patterns of gene mutations that can cause cancer

How do cars learn?

I don’t want to confuse you, so I’ll just say that machines learn by finding similar data in large data sets. The more data is transferred to the machine, the “smarter” it becomes.

Not all data is the same. Imagine that you are a pirate and your goal is to find a treasure somewhere on the island. In order to do this, you will need a large amount of information. Like the data, this information can lead you either in the right direction or in the wrong direction. The more accurate the information / data obtained, the greater the chance of success. Therefore, it is important to take into account the type of data with which it will be trained.

However, after a sufficient amount of data, the machine can make predictions. Machines can predict the future until the future is very different from the past.

Types of machine learning

There are three main categories of machine learning:

Supervised learning : A machine learns from selected data. Usually, data is marked by people.

Unsupervised learning : The machine does not learn from selected data. The point is that among the data there is no "correct" answer, the machine must find a relationship between the objects.

Reinforcement learning : The machine learns through a reward-based system.

1. Supervised learning

Training with a teacher is the most common and studied type of machine learning, because it is easier to train a machine with selected data. Depending on what you want to predict, learning with a teacher can be used to solve two types of problems: the regression task and the classification task.

Regression task:

If you want to predict continuous values, for example, try to predict the cost of a house or the weather outside in degrees, use regression. This type of task has no definite limit on the values, since the value can be any number without restrictions.

Classification task:

If you need to predict discrete values, for example, categorize something into categories, use the classification. The question “Will a person make a purchase” has an answer that falls into two specific categories: yes or no. The number of valid answers of course.

2. Unsupervised learning

Since machines do not have marked data for training, the purpose of machine learning without a teacher is the discovery of patterns in the data and their grouping.
Learning without a teacher solves 2 types of problems: the problem of clustering and the task of searching for association rules.

Clustering Objective:

Learning without a teacher solves the problem of clustering by finding similarities in the data. If there is a common cluster or group, the algorithm will classify them in a certain form. An example of this would be grouping customers based on their past purchases.

The task of finding associative rules:

Teaching without a teacher solves this problem by trying to understand the rules and the meaning of different groups. A striking example is the search for the relationship between customer purchases. Stores can find out which products were purchased together and use this information for sales. One study found that there is a close relationship between the purchase of beer and diapers. It turned out that men who went out to buy diapers for their children also tend to buy beer for themselves.

3. Reinforcement learning

This type of machine learning requires the use of a reward / fine system. The goal is to reward the car when it learns correctly, and to punish the car when it learns wrong.

Examples of reinforcement learning

- Learning machines how to learn to play ( Chess, Go)
- Super Mario , how to learn and play

We told about the basics of machine learning, the topic is interesting and promising, so take the time to learn more.

Translation of A Beginner's Guide to Machine Learning article :)

Also popular now: