Before explaining the reasons why we could be deceived by the affirmation that mathematics is not so fundamental in automatic learning, I want to investigate a little about the conceptualization of what automatic learning is.
The way we can say that we have really learned something is when we educate ourselves about some topic and test it and provide solutions. Many times that learning requires time, where we invest sacrifice until we obtain an experience that leads us to be able to affirm that we have learned, all this leads us to conclude that under this process it is difficult to affirm that the learning can be given in an automatic way, since a series of processes are required to be able to get to acquire a learning.
However, this form of learning could be said to occur in humans, since the same human being has given himself the task of achieving in machines through artificial intelligence that these can learn automatically, that is, without the need to go through all those steps to achieve learning.
Somehow computer experts have managed to enter an experimental and scientific field within computer science to achieve a kind of learning of machines and computers by means of algorithms so that they can provide solutions to any problem without the need for human intervention.
It would be illogical to think that a process as complex as that of achieving that machines, computers and other computer elements can acquire an automatic learning does not have to use mathematics, and it is that not everything ends there, to achieve that machines can be self-sustainable and through a learning system they can solve any problem that may arise requires complex mathematical knowledge.
Automatic learning is an area that is fashionable for many people who want to learn about this topic, however many of these people when they decide to learn more deeply are with a gallery of books on programming that can be useful in that learning, but when they go and go further to find the pillars that underpin this part of computing are achieved with the dreaded mathematics, which perhaps many never thought they would run into and less in an abstract and deep category as for this case represent.
The important thing for this case, my reader, that you want to enter the world of automatic learning is that you focus on a reality where you know that mathematics really is an essential part in the formation of automatic learning, so that you have an idea of how important it is that I am going to list the mathematical areas in which you should focus your knowledge in case you decide to expand your knowledge in automatic learning:
1] Vector calculus: this part is like infinitesimal calculus, i.e. the one that deals with limits, derivatives and integrals but in this case focused on the vector part.
2] Statistics and probability: this is the part of math that I think is everywhere, from the statistics that are kept in sports, such as the batting percentage in baseball, the number of triple hits scored in basketball by a player, the probability that someone will flip a coin to get a head or a stamp. In short, statistics are very common, and in the case of machine learning, it would not be left unattended.
3] Algorithms and optimization: this part is still less mathematical although the algorithms have to do with programming, flowcharts, programming languages, binary code, among other concepts, perhaps where we can see a little more classical mathematics is in the part of optimization.
4] Linear Algebra: In this part of mathematics I believe that not only those who want to know about machine learning but also future computer engineering professionals and computer graduates should learn about linear algebra, where matrices, vectors, systems of equations and vector transformations are the order of the day.
Since not all of this mathematics is so important all at once, here is a pie chart where you can see which of the areas I explained above is most relevant for getting into the world of machine learning:
It is not surprising that linear algebra and statistics are the most important areas due to the above mentioned, however, infinitesimal calculus together with vector calculus occupies 25%, emphasizing that the calculation of derivatives and integrals are not so complex if you decide to study and understand it with some care and dedication, In my university years it was one of the areas of science that I studied and learned the most, even today you can visit my blog and see that I am carrying a thematic series referred to audiovisuals where I explain exercises of derivatives and integrals.
Finally I want to add that this post is not meant to discourage you if you thought that this part of computer science had nothing to do with mathematics, on the contrary, now that you already know that you need to know and know about these branches of mathematics you should only try to keep things going along the course under their fair measure and put in mind those who pay for this type of services.
In that it is of no use to apply different techniques aimed at machine learning if there is no basis based on the mathematical areas mentioned above. And machine learning goes far beyond giving a lever and putting a machine to work to operate, necessary the design, the engineering that exists within this area, and that the more mathematical base there is in the equipment behind the analysis, the greater security there will be that what we have done within the order of application of machine learning so it will give the better result.