Home

Duodécimo Machine Learning

Introduction

Machine learning, also called automatic learning, according to Arthur Samuel in 1959, gives "computers the ability to learn without being explicitly programmed." 1

Andrew NG seems to rather saying without being apparently explicitaly programmed. [2] (https://pt.coursera.org/learn/machine-learning), 3

Take a little child, for example. One points objects to them, and says things like: - "Look, a dog!", and then, - "Look, a car!", and so on. A little while further, the child begins to mention the objects when they see them. The child learned the name of the objects and is able to classify the objects to a certain extent.

Some aspects of the learning process must be observed:

• One just shows the object and labels it.
• No particular characteristics of the object must be tought (a car has whells, a dog has eys and mouth, ...).
• The child is able to recognize up to some extent different objects of the same class (several branches of cars, with different color, size and shape, several kinds of dogs, races are retty different).

Computer codes are sets of instructions. The instructions are very strict, they are operations applyed on numerical data, like, - "add those and put it there", - "compare those and decide what to do based on the result", - "repeat this operations until ...", and so on. It is an easy task to teach to a child how to recocnize a car, one just points and label. But maybe it is not so easy to teach a computer to do the same thing expressing specific computer language instructions. This will be discussed in a more profound way further in our classes.