How does machine learning work – Machine learning or machine learning is not just interesting for mathematics and for IT companies like Google or Microsoft. But intelligence also has an immediate effect on web marketing. In the following paragraphs, we will see how artificial intelligence (AI) has evolved in recent years and what machine learning precisely means, and finally we will study the methods of machine learning and why entrepreneurs should adhere to it today. Has automatic learning systems.
Artificial intelligence is an integral part of digitalization, which has lastingly changed our society. What was science fiction a few years ago is now a reality. We speak with computers, our phones orient us and show us the shortest route, our watches understand if we have moved during the day.
The history of the machine learning system
Google And Facebook use machine learning to better understand users and provide more functionality. Facebook’s DeepFace can even now identify faces on images with a 97 percent success rate. Additionally, the giant search engine has considerably improved the speech recognition of the Android operating system, the hunt for photos on Google+ and the video recommendations on YouTube through its GoogleBrain project .Robots and automatons are a source of interest for many centuries. Already intelligence was dealt with by the writers of the romantic period and even today, we remain fascinated by robots, whether in movies, books or video games. The relation of the human being to the thinking system has always oscillated between fascination and fear . However, machine learning’s progress didn’t begin until the 1950s, at a time when computers were still in their infancy and where intelligence could only make you dream. During the two previous centuries, theorists like Thomas Bayes, Adrien Marie Legendre and Pierre-Simon Laplace had already laid the foundations for research, but we have to await the work ofAlan Turing to talk specifically about machine learning.
Robots And automatons are a source of interest for centuries. Already artificial intelligence was dealt with by the authors of the period and we remain fascinated by robots, whether in movies, books or video games, even today. The relation of the human being into the thinking machine has always oscillated between fascination and fear . However, machine learning’s progress did not begin until the 1950s, at a time where artificial intelligence could make you dream and when computers were in their infancy. During the two previous centuries, theorists such as Thomas Bayes, Adrien Marie Legendre and Pierre-Simon Laplace had already laid the foundations for research, but we must wait for the job ofAlan Turing to speak specifically about machine learning.
In 1950, It is a sort of game in it imitates conversation. If the person is not able to identify which of his interlocutors is a machine, we can then consider the computer has passed the test successfully. Two decades later, Arthur Samuel developed while improving with each game, a computer that could play checkers. So that the program had the capacity to learn. Finally, in 1957, Frank Rosenblatt developed Perzeptron, a first learning algorithm, it’s an artificial neural network.
From then On, scientists started to entrust tests, machines controlling them more or less well to their computers. Thus, he participated in the famous TV show”Jeopardy! Which had a powerful effect on the media as Watson won the round. (This event is very reminiscent of the 1997 chess contest between world champion Garri Kasparov and another IBM computer: the Deep Blue.
In 1950, It is a sort of game in it imitates human conversation. If the man isn’t able to identify which of his interlocutors is a machine, we can then consider the computer has passed the test. Two decades later, while improving with each game, a computer that could play checkers was developed by Arthur Samuel. Hence the program had the ability to learn.
From then On, scientists started to entrust their computers with tests, machines controlling them more or less well. Thus, he even participated in the famous TV show”Jeopardy! Which had a powerful effect on the media as Watson won the round. (This event is extremely reminiscent of the 1997 chess contest between world champion Garri Kasparov and another IBM computer: the Deep Blue.
Google And Facebook use provide more and machine learning to better understand users functionality. Faces on images with a 97 percent success rate. In addition The speech recognition of the has improved Android operating system, the hunt for photos on the video and Google + Recommendations on YouTube via its GoogleBrain job .
Understand What is Machine Learning
In Principle, computers, machines and programs only work the way you have configured them”if case A happens, activate B”. Our expectations for computer systems are increasing and the programs can’t foresee every conceivable situation and impose a solution. But algorithms must be available to allow apps to learn. This means that it can make relationships and that it should first be fed with data.In the Context of the machine learning system, there are related conditions that Need to be known in order to understand machine’s principle learning.
Artificial Intelligence – To create machines capable of behaving like human beings: in fact the computers and robots are expected to analyze their surroundings and so make the best decision possible. Robots must behave according to our standards. Today, AI can’t simulate the entire human being (especially emotional intelligence). Instead, partial aspects are isolated in order to deal with precise tasks. This is what is commonly referred to as weak artificial intelligence (weak AI).
Neural network – A branch Of research on artificial intelligence, neuro informatics is also trying to further design computers depending on the brain model. It confined to their own modes of operation and believes systems as abstract, which is to say liberated from their properties. Artificial neural networks are primarily abstract mathematical methods. A neural network (mathematical algorithms or functions ) is assembled like a human mind, and can cope with complex tasks. The chains involving neurons vary in power and can adapt to problems.
Big Data – The term Big Data or big data, only writes a huge data set Which reaches a quantity such that it exceeds the capacities of analysis, we talk then of Big Data. The growing media coverage of big data in recent years is a result of the origin of this data: in reality, oftentimes, the flow of information is created from user data (interests, profiles, personal information ) collected by companies such as Google, Amazon or Facebook so as to tailor the supply more exactly to clients. Volumes of information can be evaluated by traditional computer systems software can only find what the user is currently looking for. This is the reason why we now require machine learning systems that allow discovery and realization of interrelations formerly unknown.
Data-Mining – The data mining is information analysis of Big Data. Indeed, collecting alone is not of amazing value. You have To extract the relevant characteristics and evaluate them. Data Mining is distinguished from machine learning by the fact that it is mainly While the latter searches concerned with the use of models that were recognized For models.
Different machine learning methods
Basically, The algorithms are extremely different. Examples are brought by supervised learning, to the system, such as a database. Developers specify the value of data, for instance, if it belongs to category A or B. The machine learning system brings conclusions, recognizes patterns, and can handle data. The objective is to reduce the error rate.
A known example You can correct it manually, if an error is made by the system and its calculations will be adjusted by the filter. The software thus gets better outcomes.
The unsupervised Studying , ie unsupervised learning, eliminates the teacher, who in supervised learning, always indicate what goes and provide opinions on the autonomous decisions of the machine. The program here tries to comprehend the patterns. It can use clustering (partitioning of data), for example: a component is selected from the number of data, examined because of its attributes and then compared with those already examined. The object will be added to it, if it has analyzed equivalent components. If not, then it is stored.
Systems Based on learning are implemented in neural networks. Examples of applications can be seen in network security: a machine learning system detects abnormal behavior. For example, as a cyber attack can’t be attributed to a known set, the program report a problem, alarming the consumer and can then detect the danger.
In While the first method is relatively straightforward, with fairly superficial results, deep learning (or deep learning) is more challenging to understand. This is very complex information, since it’s natural information, such as that which occurs during speech, writing or facial recognition. Because it is difficult to enter 21, natural data is simple for humans to process, but not for a machine.
Deep A network of neurons and learning are closely linked. How a network is formed can be described as profound learning. It is called deep learning because the neural network is organized into several hierarchical levels. They record the information, start their investigation and deliver their results to the neural node. At the end, the increasingly information reaches the initial level and the network provides a value.
To Illustrate and better understand deep learning, we can use such as Google Image search. The network, which is behind the search algorithm, only provides images that show cats when searching with the phrase”cat”. Because Google’s machine learning system can recognize objects it works.
By Surfing the layers, the filter selects only the information necessary until it Can decide the image, for example a cat. During the training phase, the developers provide a class for each Image that the system can find out. If the machine produces false Results the programmers can then Adapt the neurons. Like our mind, they have different Weights and thresholds that can be corrected in a machine learning system