Content-addressable memory

An associative memory or content addressable memory ( Content Addressable Memory Data Sheet, CAM) is a memory shape, is carried out in association with the content, to access individual memory contents.

Another description would be that the access to a memory content by entering a memory value and not a memory address is set. However, the special nature of an associative memory is not due to the input of whatever interpretable ( value / address) string, but in the first independent on the particular access structure organization - best understand the question: ' What is to not described locations? ' While addressing regularly are with explicit addresses a list or table structure again, in which the defined from memory locations are occupied first with a ' blank ' value are incurred in the idealized associative memory at all only there memory locations for which previously a write operation has taken place.

In 1943 Konrad Zuse suggested before this storage form, but they could only be realized with the development of semiconductor technology. A typical use of this type of memory are fast cache memory. And memory tables which are accessed frequently may be implemented as a content addressable memory. For example, the table of MAC addresses is designed as a CAM in high-quality network switches. Another field of application are data structures in Artificial Intelligence. So also the human memory works with associations. Humans for example, connects with specific objects memories of certain experiences. A Ternary Content Addressable Memory ( TCAM ) operates at a third logic value, do not care, in order to access the memory contents.

Today, this form of storage is largely replaced by hashing techniques, which work with conventional memory and thus are much less expensive to implement.

For computer systems, the realization of large associative memories is difficult, as well as to perform " fuzzy" inputs to a desired result ( fuzzy search). One can look at this problem in general so that an input vector is to provide an output vector if it has enough resemblance to a pattern vector. To realize such functions, will make use of neural networks.

A very simple implementation of the allocation as possible, for example by one of the simplest neural networks, the single-layer perceptron.

If you add several associative memory together so that in them except data and programs can be stored and executed, creates an associative machine.