You get a bonus - 1 coin for daily activity. Now you have 1 coin

Nearest Neighbor Method for Static Pattern Recognition

Lecture



Here the idea is that around a recognizable object Nearest Neighbor Method for Static Pattern Recognition volume cell is built Nearest Neighbor Method for Static Pattern Recognition . At the same time, an unknown object belongs to that image, the number of training representatives of which in the constructed cell turned out to be the majority. If we use statistical terminology, then the number of image objects Nearest Neighbor Method for Static Pattern Recognition trapped in this cell characterizes the estimate of the volume averaged Nearest Neighbor Method for Static Pattern Recognition probability density Nearest Neighbor Method for Static Pattern Recognition .

To assess the averaged Nearest Neighbor Method for Static Pattern Recognition need to solve the question of the relationship between volume Nearest Neighbor Method for Static Pattern Recognition cells and the number of objects of one or another class (image) that fell into this cell. It is reasonable to assume that the smaller Nearest Neighbor Method for Static Pattern Recognition the more subtly be characterized Nearest Neighbor Method for Static Pattern Recognition . But at the same time, the fewer objects will fall into the cell of interest, and therefore, the less reliable the estimate Nearest Neighbor Method for Static Pattern Recognition . With an excessive increase Nearest Neighbor Method for Static Pattern Recognition reliability of an assessment increases Nearest Neighbor Method for Static Pattern Recognition , but the subtleties of its description are lost due to averaging over too large a volume, which can lead to negative consequences (an increase in the probability of recognition errors). With a small amount of training sample Nearest Neighbor Method for Static Pattern Recognition It is advisable to take extremely large, but to ensure that within the cell density Nearest Neighbor Method for Static Pattern Recognition little changed. Then their averaging over a large volume is not very dangerous. Thus, it may well happen that the cell volume relevant for one value Nearest Neighbor Method for Static Pattern Recognition may not be suitable for other cases.

The following procedure is proposed (for now, we will not take into account the belonging of an object to a particular image).

In order to evaluate Nearest Neighbor Method for Static Pattern Recognition based on a training set containing Nearest Neighbor Method for Static Pattern Recognition objects, center the cell around Nearest Neighbor Method for Static Pattern Recognition and increase its volume as long as it does not contain Nearest Neighbor Method for Static Pattern Recognition objects where Nearest Neighbor Method for Static Pattern Recognition there is some function from Nearest Neighbor Method for Static Pattern Recognition . These Nearest Neighbor Method for Static Pattern Recognition objects will be closest neighbors Nearest Neighbor Method for Static Pattern Recognition . Probability Nearest Neighbor Method for Static Pattern Recognition vector hits Nearest Neighbor Method for Static Pattern Recognition to the area Nearest Neighbor Method for Static Pattern Recognition determined by the expression Nearest Neighbor Method for Static Pattern Recognition .

This is a smoothed (averaged) density distribution. Nearest Neighbor Method for Static Pattern Recognition . If you take a sample of Nearest Neighbor Method for Static Pattern Recognition objects (by a simple random selection from the general population), Nearest Neighbor Method for Static Pattern Recognition of them will be inside the area Nearest Neighbor Method for Static Pattern Recognition . Probability of hitting Nearest Neighbor Method for Static Pattern Recognition of Nearest Neighbor Method for Static Pattern Recognition objects in Nearest Neighbor Method for Static Pattern Recognition described by a binomial law having a pronounced maximum around the mean Nearest Neighbor Method for Static Pattern Recognition . Wherein Nearest Neighbor Method for Static Pattern Recognition is a good estimate for Nearest Neighbor Method for Static Pattern Recognition .

If we now assume that Nearest Neighbor Method for Static Pattern Recognition so small that Nearest Neighbor Method for Static Pattern Recognition inside it changes slightly, then

Nearest Neighbor Method for Static Pattern Recognition ,

Where Nearest Neighbor Method for Static Pattern Recognition - area volume Nearest Neighbor Method for Static Pattern Recognition , Nearest Neighbor Method for Static Pattern Recognition - point inside Nearest Neighbor Method for Static Pattern Recognition .

Then Nearest Neighbor Method for Static Pattern Recognition . But Nearest Neighbor Method for Static Pattern Recognition , Consequently, Nearest Neighbor Method for Static Pattern Recognition .

So, the assessment Nearest Neighbor Method for Static Pattern Recognition density Nearest Neighbor Method for Static Pattern Recognition is the value

Nearest Neighbor Method for Static Pattern Recognition . (*)

Without proof we give the statement that the conditions

Nearest Neighbor Method for Static Pattern Recognition and Nearest Neighbor Method for Static Pattern Recognition (**)

are necessary and sufficient for convergence Nearest Neighbor Method for Static Pattern Recognition to Nearest Neighbor Method for Static Pattern Recognition in probability at all points where the density Nearest Neighbor Method for Static Pattern Recognition continuous.

This condition is satisfied, for example, Nearest Neighbor Method for Static Pattern Recognition .

Now we will take into account the belonging of objects to one or another image and try to estimate the posterior probabilities of the images. Nearest Neighbor Method for Static Pattern Recognition

Suppose we place a volume cell Nearest Neighbor Method for Static Pattern Recognition around Nearest Neighbor Method for Static Pattern Recognition and grab the sample with the number of objects Nearest Neighbor Method for Static Pattern Recognition , Nearest Neighbor Method for Static Pattern Recognition of which belong to the image Nearest Neighbor Method for Static Pattern Recognition . Then according to the formula Nearest Neighbor Method for Static Pattern Recognition estimation of joint probability Nearest Neighbor Method for Static Pattern Recognition there will be a magnitude

Nearest Neighbor Method for Static Pattern Recognition ,

but

Nearest Neighbor Method for Static Pattern Recognition .

Thus, the posterior probability Nearest Neighbor Method for Static Pattern Recognition estimated as the fraction of the sample in the cell related to Nearest Neighbor Method for Static Pattern Recognition . To minimize the error level, you need an object with coordinates Nearest Neighbor Method for Static Pattern Recognition attributed to the class (image), the number of objects of the training sample of which is maximum in the cell. With Nearest Neighbor Method for Static Pattern Recognition such a rule is Bayesian, that is, it provides a theoretical minimum of the probability of recognition errors (of course, the conditions Nearest Neighbor Method for Static Pattern Recognition ).

See also


Comments


To leave a comment
If you have any suggestion, idea, thanks or comment, feel free to write. We really value feedback and are glad to hear your opinion.
To reply

Pattern recognition

Terms: Pattern recognition