Today, I'll go over the basic idea of machine learning. We'll boil machine learning to its basics without getting too technical and then explain why machine learning is useful.
Functions
To understand machine learning, we will first need to understand functions. Functions are basically mappings from one space to another space. To explain this concept more simply, a function is used to transform some input (within the input space) to some output (within the output space). In algebra, we take some can express an input x in terms of an output y:
y = 3x
What we are doing is describing the y space and its relationship with the x space for all valid inputs (Typically the x space is called the domain and the y space the range, but these terms are not really that important to understanding the idea). We can take a specific point in the x space like 5 and map it to the y space where it is 15. Functions are useful as they explain relationships between different variables. For example, one space could be the number of people online and another space could be the price of bitcoin. A function could explain to us the relationship between the two.
Hidden Functions
But why are functions useful in understanding machine learning? Well most of the time we don't have a ready made function describing the relationship between two things that we want to know. Lets say we want a function that takes the number of people online and maps it to the price of bitcoin. How do we find that function? Well if we were to do it by intuition, we would need some sort of prior experience to help us out. Thus, we need snapshots of the number of people online and the price of bitcoin at different times. Using these snapshots, we can build a model to estimate the hidden function. A model which will describe the relationship between our two different variables.
Learning From Experience
Machine learning is built upon the intuition that people learn from prior experience. Thus, data points serve as the knowledge that machine learning algorithms use to produce a model. A model is basically an approximation or hypothesis of what the algorithm thinks the function is. We use the word hypothesis because we can never really know how accurate the model is with the function being hidden. We do have methods of approximating accuracy, which we won't go over this time.
A possible model for some hidden function
Summary
To summarize to this post, we use machine learning algorithms to develop models which are guesses to what hidden function we are trying to discover. This hidden function is some underlying relationship between one space or set of variables x and another space y. Machine Learning is useful as almost all aspects of our life can be described using variables, and we can discover useful relationships, predict future outcomes, and find new ways to describe all sorts of different data from the functions we discover. This idea of function discovery via machine learning is allowing us to create driverless vehicles, make financial predictions, give computers the ability to see what is around them and much more!
Thanks for reading and hope this demystifies machine learning for you. I'll probably go over a few more introductory topics in the future and can go more in depth or cover specific algorithms if anyone is interested. Feel free to upvote, follow, and resteem if you found any of the information here useful.