I would like to know when one can apply the nearest neighbor algorithm, like
image processing, robotic, DNA sequence, etc.
Nearest neighbor algorithms can be used to improve slightly a given feasible solution to a complex problem it can be as well joined with global search algorithms like genetic algorithms.
Nearest neighbor algorithms is also can be use with the image processing. This method simplify and determine the nearest neighbor pixels. Please consider the intensity value of the pixels.
The Nearest Neighbohr Classifier is a supervised machine learning algorithm to solve classification Problems. As far as Training and classification complexities are concerned, this algorithm is far simple when compared to other classification algorithms such as neural Networks. Consequently, it can be easily be applied on low-computing and low-battery-power devices. However the Features should ideally be numerical, so that the distances between instances can be computed and the distance(Euclidian, Chebychev...) as we commonly understand it, should be relevant to the Notion of class. Don't Forget that distance has more to do with the Notion of Cluster than the Notion of class. That is, two data Points can be very closed and not be Elements of the same class. EXAMPLE; If we give you as data set the positions of all humans on the Earth and ask you to Cluster them, then using the Notion of distance , you will probably end with the map of the world where clusers will be countries/continents separated by seas..
Now suppose that a labeled subset of this data set is given to you, that is the Position of a Person together with the Person nationality(class), Using the Nearest Neighbohr classifier here fundamentally presumes that there is a high correlation between one's nationality and the place/country where one lives. That was true 20 ore more years ago. But GLOBALISATION (The world is one) changed and is changing this view. I am Cameroonian , I study in Germany and all my friends are chinese students. So Nearest Neighbohr Classifier may not be suitable to this case because distance is not so relevant to the Notion of class.
it is for classifying the new data with existing clusters of data by using different methods to measure the distance between the new data and the clusters and there are an updated version of it called "KNN" or K nearest neighbor
Knn is also used in social networks. As example, it can be used to identify and connect similar users in a users ranking service.
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Artificial Intelligence