Basic Introduction to Maching Learning
The machine learning is the subfield of computer science and a branch of artificial intelligence but in order to understand machine learning, it is first important to know that it is artificial intelligence and algorithms since they are related areas and are part of machine learning.
Artificial intelligence
Artificial intelligence is the term given for the tasks that we must perform and which are necessary for human intelligence to be executed. Its beginnings are in computing and dates from the 60s. It was created to solve simple and basic human tasks, but which, at that time, was a challenge for the primitive computers of the time.
The term artificial intelligence is applied to machines or technological devices when they imitate cognitive functions and capacities related to human beings, such as, for example, perceiving, reasoning, learning and solving problems. Through the information that is entered into the machine or that it is capable of collecting on its own, you can learn from such data and use it to achieve specific tasks and goals.
Examples
A simple example would be a thermostat in the home, which turns on a cooling fan based on the temperature.
Home automation allows us to control our lights or electrical appliances in the house according to our preferences for both use or schedules, allowing us to interact with various objects in our house without having to turn them on or off, among them are the lights in our house, systems irrigation, music, alarms, doors, among others.
Algorithms
They are a set of instructions or rules previously defined to solve a problem, be it computing, processing data or performing tasks and activities. In everyday life we use algorithms unconsciously every day.
A simple example would be the steps we take to go to eat:
Buy or eat
1. Wash our hands
2. Serve the food
3. Sit at the table
4. eat
Another example would be to find the phone number of a person in a physical address book that is arranged alphabetically, the algorithm for this problem would be:
1. First search the contact list or phone book.
2. Open it in the middle and check if the letter of the page we are on is higher or lower than the one we are looking for.
3. If it is larger, we split the book in half once more and if it is smaller we do the same for the other end of the book and re-analyze the previous condition (find if the letter in the middle matches the name we need) and repeat the process of dividing into halves until we find the letter that matches the name.
4. When we find the letter with which the name begins, we look on the same page from top to bottom for the name and once we look at the name, we will find your phone next to it.
5. We review the number and once we no longer need it, we close the book.
What is machine learning?
Machine learning is automatic learning itself, the part of Artificial Intelligence dedicated to giving the correct answer or solution to a question or problem. The magic of machine learning is that this answer is given without the choice mechanism being previously programmed.
The system learns by itself to give such an answer or to solve a problem, based on previous experiences it has had with those tasks (input and output data).
In technology, machine learning is a branch of artificial intelligence that creates systems that learn automatically, that is, to be able to identify complex patterns in millions of stored data, machines programmed in machine learning are capable of predicting future behaviors taking into account the analysis of previously stored data and this autonomously, without the need for human intervention for its operation.
Its branch focused on neural networks is called deep learning.
Deep learning
It is a more advanced learning since it makes use of neural networks simulated by machines. It is a deeper approach to Machine learning and its objective is to build knowledge through the simulation of the human brain, on a more basic scale.
Different machine learning algorithms.
Supervised learning:
The supervised is the one who has been previously told what to analyze or look for and the program is limited to the conditions that have been taught from the beginning in its instructions. It allows you to make decisions or make predictions. An example is a spam detector in our email.
Unsupervised learning:
Unsupervised learning for its part, the algorithm that it has is modified by itself and does not contain prior knowledge of any specific operation, so it takes factors indistinctly that have neither order nor name, yet it analyzes the information according to its programming.
Most of the machine learning that is used in companies is supervised and consists of deep analysis and they are programs so large that they work through networks in very powerful machines, which can perform “mathematical” operations with great capacity and agility.
Reinforcement learning:
The goal is for the algorithm to learn from its own experience. That is, the best decision you can make correct decisions based on the trial and error process. Currently, it is used to enable facial recognition, perform medical diagnoses, or classify DNA sequences.
Machine learning in companies
Buyers and consumers:
A telephone company wants to know which customer is in danger of canceling their services with the company in order to take any action to keep the customer or at least reduce the risk of losing the customer and the customer decides to contract with the competition.
What the company does is analyze the extensive information it has on all its customers, an immense amount with data from contracted plans, daily consumption services, specific use of each of its contracted services, to predict when they are about to lose a business. client and be able to take actions to avoid it. A person or a group of people is unable to analyze so much information and even more to find a “predictable” behavior in the data, but algorithms are capable of checking behavior patterns. With maching learning they can be more proactive with the information, all customer data is stored in an orderly manner in databases (programs capable of storing large amounts of data) to be able to analyze and exploit them, predicting future behaviors both beneficial and detrimental to the objectives of the business.
Business security:
Through the recognition of images it is possible to detect the faces of the members of a company, since in the system the faces with photos and videos are stored and it is possible to identify who enters and leaves a company with all the related information, schedules and others required information.
Time optimization:
The system will determine when it is time to do a specific task for each objective, for example, stop production, increase production, investments, reduction of budgets, etc.
Customer service by voice and text:
Through frequent patterns of use of phrases and conjugation of words, it is possible to generate coherent and adequate answers for each question generated by the user, making repetitive processes in some frequently asked questions more agile and with less work.
Information security and anti-fraud:
It is possible to analyze the behavior of objects and people on a perimeter or terrain, we can through patterns identify attacks and alerts before they happen.
It is also possible to detect that users are fraudulent in e-commerce sites through their profiles, data and access methods and to reduce the percentage of thefts through cards, credentials and others.
Other applications
Medicine:
Researchers at the Massachusetts Institute of Technology (MIT) are already using machine learning to detect breast cancer early, which is crucial because early detection of breast cancer increases the chances of a cure. It can also be used to detect pneumonia and diseases of the retina that can cause blindness.
Neuro-linguistic programming:
In our cell phones or mobile devices they are already equipped with virtual assistants such as Alexa or Siri that can instantly translate from one language to another, recognize the user’s voice and even analyze their state of mind or feelings. On the other hand, the NLP also performs other tasks such as helping managers, lawyers and administrative personnel to organize large volumes of information related to their requirements or searches.
Search:
Current search engines are able to identify your content preferences regarding the information you are looking for and the recurring topics you frequent on the internet and show you suggestions for topics that may interest you related to your searches
Social networks: some social networks use Maching learning to greatly reduce published spam and eliminate fake news as disallowed content that they automatically block.
Smart vehicles:
In the next few years it is planned to see smart vehicles on the roads. Thanks to Maching learning, these future vehicles may have an internal configuration to adjust the temperature, music, backrest inclination, etc. According to the driver’s preferences and even, move the steering wheel alone to react to the environment by itself.
References:
https://www.iberdrola.com/innovacion/machine-learning-aprendizaje-automatico
https://andro4all.com/2018/10/machine-learning-que-es-para-que-sirve
