Differences and relations Classification and Regression
contact
Contact return to nature and classification is to be established mapping relation
\ (f (x) \ rightarrow y, x \ in A, y \ in B \)
the difference
The fundamental difference between regression and classification is whether output space is a metric space
For regression problems , which is the output of a metric space space B, the so-called "quantitative" . That is, the regression output space defines a metric \ (d = F \ left ( y_ {true}, y_ {pred} \ right) \) to measure the "size error" between the output value and the true value. For example: predicting the price of a bottle of 700 ml cola (real price of $ 5) a 6-membered, the error is 1; prediction which is $ 7 2 error. Both predictions are not the same, there is a measure defined to measure this "different". (Mean square error so with such error function).
For classification problems , the output is not a metric space B space, the so-called "qualitative." In other words, classification, only the classification of "right" and "wrong" parts, as when an error is assigned to the Class 5 Class 6, or Class 7, and there is no difference, are in error counter +1
In an overview of FIG.
to sum up
Take for example the support vector machine, classification and regression problems must be found in accordance with the training sample a real-valued function g (x).
Regression requirements is: Given a new model, the output of its corresponding inference based on the training set y (real number) is how much. I.e. using y = g (x) to infer any input value x corresponding output.
Classification problem is: given a new pattern, based on the training set to infer the category corresponding to it (such as: + 1, -1). I.e. using y = sign (g (x) ) a category corresponding to the input x to any inference.
In summary, regression and classification problems, like nature, different only in that the value range of their output.
Classification problem, the output of only two values ;
and in the regression, the output could be any real number
A fun chestnuts
by pictures of people judge a person is not fat? (This is a classification problem: is not fat or fat)
by pictures of people judge a man weigh? (This is a regression problem: high Wenxin looks 100kg)