Machine Learning
Machine learning, what is machine learning. I think Shangxue Tang teacher training is very good, is to introduce the word. ------Personification.
Analog is the human way of thinking.
Teacher cited an example:
And girlfriends dating, first date, agreed to 19 o'clock, but only to the girlfriend 7:10.
The second time, date, also agreed to the night at seven o'clock, but the girlfriend was still 7:10 to.
So it third time, on the gay experience may be pre-dating by two, ten minutes late girlfriend made the prediction is the probability. Also ten minutes later to go out in ten minutes you can read a book, listen to songs or something. Then this is the prediction of the future to make a value judgment.
It is also possible, the first two are on the way to his girlfriend run into traffic jams, road or road detour.
So if the number of dating more than 365 days a year, there are 300 days late, then the probability of the next girlfriend late for an appointment will be enormous.
Therefore, machine learning, the prediction is based on data on the basis of the extracted feature value data in the mass, by the algorithm to build a model. To determine the validity of the model by measurement data. Then change the parameters to improve the accuracy of prediction.
Machine learning courses at Stanford University, the two definitions are given.
The first is:
The field of study that gives computers the ability to learn without being explicitly programmed.
The computer will be able to learn without explicit programming of research.
The second offers a more modern definition:
"A computer program is said to learn from experience E with respect to some class of tasks T and performace measure P ,if its performance at tasks in T,as tasks in T,as measured by P, improves with experience E"
"It is said that a computer program can learn from the experience of some of the implementation of the E T P class tasks and measures, if it appears in the task T, that is T task performance, as measured by P, with the experience of E improve and improve, then it can be measured P a. "
And he cited an example:
Example:playing checkers.
Play checkers
E = experience of playing many games of checkers.
E = play a lot of disk checkers experience
T = the task of playing checkers.
T = the task of playing checkers
P = the probability that the program will win the next game.
Under the program a win probability P =
In generally, any machine learning problem can be assigned to one of two broad classifications:
All in all, any machine learning problem can be divided into two categories
Supervised learning and unsupervised learning.
Supervised learning and unsupervised learning.