Source: Deep Learning on Medium
Grandpa, Do you want to know about Machine Learning?
It was always you who taught your children and grandchildren how things work with your experience. Today, when I have the opportunity to share with you some of what I know, I want to talk about technology, especially about a subject that without you knowing its definition is already among us and perhaps you have used this technology many times.
You are fortunate, you have seen the evolution of technology almost in its entirety, you have seen how humanity, to cite an example, has gone from writing texts on typewriters to computers, to cell phones and now to dictate them to a machine and that it is this one that performs the writing automatically.
What is Artificial Intelligence?
The first thing I want you to understand is that it is Artificial Intelligence, since it is the biggest concept and we will see later how it relates to Machine Learning.
We will define artificial intelligence as the development of machines that have the objective of imitating human behavior. And at what level does the machine try to imitate human behavior? What is sought is that the machine can develop the same as a human in the same way or even more efficiently. For example, have you heard that in the near future the vehicles will be autonomous? That is to say, that they will run without drivers? Well, in this case the intention is that the vehicle can drive and transport people from one place to another as a driver would do, but by doing so autonomously, the number of accidents would be reduced.
Likewise there are many machines already doing work that was exclusive to human beings, for example at home we already have the robot that sweeps and collects the garbage from all over the house, going around all corners and without generating any accident.
As you can imagine, there are many applications for artificial intelligence, where due to the processing capacity of the machines, they can perform much more efficient tasks than human beings.
You may be wondering if we can say that it’s artificial intelligence at all that we see performing tasks for us. Well, there is a method that was created to know how much a machine can emulate the behavior of a human and do it as if it were a human itself.
The initial turing test, which is the name given to this method, consists of separating a human in one room and in the other room there can be a human or a machine and they pass papers to each other under the door carrying on a conversation. If the human in the first room cannot identify if in the other room there is a human or a machine because he was responding adequately, it means that if there is a machine, it has artificial intelligence because it is simulating the behavior of a human.
Likewise, with the example we used of the vehicles that are driven alone, if a vehicle is taken and during the trip it is not distinguished if the vehicle is autonomous or is driven by someone, it is an indication that it is a vehicle with artificial intelligence.
I hope this short explanation makes it clear to you that it is artificial intelligence and how to identify it. Now that we have this concept clear, let’s move on to machine learning.
What is machine learning?
I know you are wondering: How does the machine look so much like a human behavior and not look every time a machine but a human? The answer to that question can be found in machine learning.
Machine Learning is a branch of artificial intelligence and aims to equip machines (computers) with the ability to learn, to analyze data, to identify patterns and make decisions with minimal human intervention. Thus, the machine is provided with a large amount of data, which are processed through algorithms and this is sufficient for the machine can provide a result with a specific behavior as if it were a human making decisions with the information provided.
Do you see the relationship with Artificial Intelligence? Artificial intelligence needs machine learning so that machines learn every time how to respond to the data they obtain in one way or another.
To make clear all the concepts, let’s define what an algorithm is, because it is nothing more than a sequence or series of instructions, which represent the solution to a certain problem.
In conclusion, with the right algorithm the machine can analyze the data and have the right behavior for certain situations and just as we do that we learn from what we live, the machine can learn from the data entered and the expected results each time an event occurs.
In order for the machines to learn efficiently, the algorithms have been divided into three groups:
n this type of algorithm we have a previous knowledge of the data, we know them and we know what they are and what they are for. The process in this case implies that the data are previously tagged by humans so that the machine from them has the knowledge inherited from the tags made. For example, with the pictures you can clearly show the machine which are dogs and which are cats by labeling each picture accordingly. In other words, these are problems that we have already solved, but that will continue to arise in the future. The idea is that the computers learn from a multitude of examples, and from there they can make the rest of the calculations necessary so that we don’t have to enter any information again.
In this type of algorithm there is no previous knowledge of the data, so that they can be labeled, so it is sought in these cases are patterns that can from the data classify information and determine the characteristics themselves that may have an object to examine. Using the same example of the previous case, if enough data about the dogs and cats are provided, the machine will be able to establish that they are two different animals and that they have very different characteristics and when some characteristics of one of these are provided it will be able to classify it easily and deliver the result according to what it has been learning.
In this case, the basis of learning is reinforcement. The machine is capable of learning by trial and error in a number of different situations. Although it knows the results from the beginning, it does not know what the best decisions are for achieving them. Therefore, the algorithm progressively associates the patterns of success, to repeat them over and over again until they are perfected and possible failures are reduced to a minimum. An example for this case are the automatic drones which were learning from the errors and tests performed until they had enough information to avoid accidents during flights.
And with all this who uses machine learning?
Uses of machine learning.
oday there are many industries that are using machine learning, as you can analyze, there are many data that have been collected by each of the companies throughout its existence and with which they can make better decisions (information about customers, suppliers, machinery, etc).
To give you clarity I am going to give you only some examples of where machine learning could be used:
– In banks to detect fraud in transactions.
– In companies to predict failures in technological equipment.
– For human resources to predict which employees will be more profitable in a certain period
– To Select potential customers based on social media behaviors.
– For state-owned companies To predict urban traffic.
– For the health sector To make medical pre-diagnoses based on patient symptoms
– For the commercial sector Decide on the best time to call a customer.
As you can see, machine learning is being widely used by almost all sectors of the economy. Do you understand why you are getting promotions for that product that you used to consume frequently and that you are not buying now?
To finish grandpa, I’m going to tell you a little bit about Deep Learning.
Deep Learning carries out the Machine Learning process using an artificial neural network that is composed of a number of hierarchical levels. At the initial level of the hierarchy the network learns something simple and then sends this information to the next level. The next level takes this simple information, combines it, composes a somewhat more complex information, and passes it on to the third level, and so on.
If we take into account the example of the dog and the cat that we have used before, with deep learning we can establish more accurately if a picture is of a dog or of a cat. Initially deep learning can determine the contour of the animal and pass it to a second stage that can evaluate characteristics such as hair, height, shape, with the result obtained a third layer can evaluate the shape of the eyes, nose, mouth, legs, a fourth layer can evaluate the shape of the beard, coat and so subtly will go through layers to determine exactly that all the characteristics correspond to a particular race.
Deep learning is closer to the human perception of things and that is why it becomes so interesting in modern times.
You already have the concepts of the technology that is changing the world today and I hope I have been clear enough with everyone that you can include this information in your talks with your friends, or that you can instruct another of your grandchildren.
I love you Grandpa.