Machine learning for novices Part 1

With machine learning technologies today are confronted daily every resident of the metropolis. But not everyone knows what machine learning is really capable of.

Machine learning every day takes more and more place in our lives due to the vast range of its applications. From analysis of traffic and ending self-driving cars, more tasks is passed on to same because machine.

We sometimes even approximately do not represent, as some application based on methods of machine learning. For example, no one will be able to answer the question “Why am I today in advertising showed the site A and not B?”. The most sad in this whole situation is that most people have the wrong idea about machine learning.

Introductory

Machine learning is considered a branch of artificial intelligence, the main idea of which is that the computer is not just used pre-written algorithm, and he learned the solution of the problem.

Any workable technology of machine learning can be roughly categorized into one of three levels of availability. The first level is when it is only available to various technological giants of level of Google or IBM. The second level is when she can take advantage people the student with some knowledge. The third level is when even grandma is able to control it.

Machine learning is now at the junction of the second and third levels, whereby the rate of change in the world using this technology is growing every day.

Learning with teacher and without a teacher

A big part of machine learning tasks can be divided into supervised learning (supervised learning) and learning without a teacher (unsupervised learning). If you imagine a programmer with a whip in one hand and a lump of sugar to another, you may be confused. Under the “teacher” here refers to the idea of human intervention in data processing. In supervised learning, we have data, based on which the need to predict and some hypotheses. In unsupervised learning, we only have data properties which we want to find. Examples of the difference you will see a little clearer.

Learning with a teacher

We have information about 10 000 apartments in Moscow, and famous for the area of each apartment, number of rooms, the floor on which it is located, the area, availability of Parking, distance to the nearest metro station and so on. In addition, the cost of each apartment. Our task is to construct a model which on the basis of these characteristics will predict the price. This is a classic example of supervised learning where we have data (10 000 apartments and various parameters for each apartment, called signs) and responses (the cost of the apartment). Such a problem is called a regression task. That it is, we will talk later.

The red dots of the available data (the x — axis is the characteristic value for the y — axis the response value), blue video — built model.

Other examples: based on various health indicators to predict the presence of cancer patient. Or on the basis of email text to predict the probability that it is spam. Such tasks are classification tasks.

The classification problem. In the first picture, the objects are separated by a straight line. The second more complex curve. Note that some objects classified incorrectly. It is normal practice in classification tasks.