Concept Learning

Concept learning Concept learning is a task of inferring a Boolean-valued function from training examples of its input and output. The other way we can say, concept learning is the task of finding a hypothesis, which is consistent enough. Let’s assume there is a task or a problem, we can say, to find out which days VII come out to enjoy sports. The actor, which in the past experience, has certain features like sky condition, humidity, temperature, wind, water, and forecast. So, these are all called attributes of the day, which take on values. So many hypotheses, the problem is to find out the one which is consistent, for which we use to refer to the past history, past experience, with a function F of X. Learning task has a target concept, that is, days on which Talab VII enjoys his favorite water sport. This is what the concept target concept is. With a given set of example days, each represented by a set of attributes, the attribute for enjoy sport, that is what our problem is, which indicates whether or not VII enjoys his favorite sport on this day or not. The task here is to learn to predict the value of enjoy sport for any arbitrary day based on the values of its other attributes. Now, let’s come to the point of representing the hypothesis. Here, the task of finding an appropriate set of hypotheses for concept learning in a given training data. This hypothesis is a conjunction of constraints, which takes the form. It may have a certain specific value, like if it is water, it may be warm water or cold water—so a straight value it can take, a specific value it can take. Otherwise, it may take a “don’t care” value, that means whatever it may be, that is, there is a question mark, and no value allowed condition, where there is an empty. Water is empty here, right? So, there is no permissible value given to the attribute. Normally, the hypothesis is represented as a vector. All the attributes of the task are listed over here, like sky, air temperature, humidity, wind, water, and forecast. There are six variable attributes. The hypothesis, initially, this is a sample one only. So, the sky must be sunny, but the air temperature and humidity may be anything. The wind must be strong, and water may or may not be there. The forecast must be the same. So, this is one sample hypothesis, how it is formulated and framed. Here, the idea is satisfaction of a hypothesis by some examples. So, say the example which satisfies the hypothesis with the defined function H(X), where H(X) is equal to one if this H is true on the particular sample X. Suppose if it is not satisfied, then H(X) will be zero. So, we have to find hypotheses that best fit the examples. Also, we can reduce the learning through a search problem over a space of hypotheses. So, anyway, out of all the hypotheses We have to find the best fit through a search problem. Understood? A scenario is given: VII enjoys his favorite sport only on cold days with high humidity, right? We can represent this expression using the hypothesis. So, first, only the specific temperature is considered, which is “cold,” and humidity must also be high. So, only for two attributes, specific values are given in this example. The hypothesis is formulated by replacing the question mark with the attribute value given, while the rest will remain as question marks. The most general hypothesis always takes the form of all question marks, which means it is free to accept any value. The most specific hypothesis is all empty (Ø symbol), which means no value is allowed.