Neural Networks: Data compression and associative Memory

The ability of neural networks to identify relationships between different parameters makes it possible to express data of high dimensionality is more compact, if the data are closely linked with each other. The reverse process - the restoration of the original data set of pieces of information - called the (auto) associative memory. Associative memory can also restore the original signal / image from the noisy / invalid input. geteroassotsiativnoy Memory Solution allows for memory addressable by content. It is this property of neural networks, I want to dedicate my graduation test. The theoretical part will consist of a knowledge of the methodology that I can find, as well as related theories and other information that will be required for disclosure of the topic. The practical part will include a network (architecture still remains open to question), that input will receive some information that you want to compress, and output, respectively, the same data is already compressed. It is also likely to be implemented network performs the opposite effect. An item will udelёn so-called "associative memory". Given enough time, I will try to implement such a neural network.