Artificial neural networks and search engines
Artificial neural networks (KNN) are models which are modeled on the biological networks in a human brain. The nerve cells and neurons are connected in the brain in a dense network. While we learn something, change (synapses) connections between cells are necessary for the learning process. This adjustment we are able to learn to cope with different tasks and solve problems. With the computer this process in the form of similar models can simulate. Artificial neural networks imitate to the nature of the brain to organize by learning processes. In principle they can be used for each task, which is, to identify links between “fuzzy” patterns.
KNN consist of a group of cells that are arranged in multiple layers.
There are the input cells that serve type of problems. The so-called hidden cells are responsible for processing the task. Finally, there is still the output cells that output solution from learning and processing.
A simple example to describe neural networks represents as the character recognition. Should the “number 1″ be identified, no matter how it is on a pixel image. Where can the number twisted 1″, little contrasting, be shown even greater and even less, and so on. Output neurons indicate whether it is the “number is 1″ or different character.”
For best results in the neural network, fed with training data. That is one the neuronal network passes, if the “number 1″ or another character is.
After a certain number of training passes the neural network has hardly error and can be used now to give correct results based on different pixel images.
What has to do it now with search engines? For a long time in numerous research projects examined and used skills artificial networks. Why then not for the search engine algorithm? The desire of the search engine providers is to provide the search engine users of high-quality, semantic content. But too often the user’s search inputs are “fuzzy”. Incorrect or inaccurate search terms would such examples. The search engine must first know what word types it is: noun, verb, adverb, adjective,… In which semantic environment are these concepts and how are the words to each other related. The more keyword information, the more information be . And to the search engine how you can use now artificial neural networks for search engines? The Internet user best detects whether the requested content really is who he has search (training set). Whether it is to the correct result is, you could for example by submitting voice or voting of seeking (such as Wikia search) or set by the last visited result of search engine log data. You could train so neural networks on keywords get always better and higher quality results. But requires many machines (e.g. Google computer farms). Perhaps Google therefore logs the Internet usage data of its users. As often happens, the search engine results are more accurate if the use of neural networks is also included.
P.S.: I am in my online marketing and marketing software blog used the tool June before the simulation of artificial neural networks. Just call if you are interested and test