|
Content-addressable memories
are nicely motivated in this chapter and illustrated in terms of two important types of neural
networks: Hopfield
and Kanerva memories.
After conceptually introducing both these memories,
Hopfield networks are presented and discussed in terms of stability. This type of content-addressable
memory is illustrated through a simple example involving the retrieval of entries in a phone book.
In the next section in this chapter Kanerva memories, which unlikely
Hopfield networks can be
auto or heteroassociative, are motivated and introduced in a particularly careful fashion.
The implementation and performance of these memories are also nicely discussed and illustrated.
|
|
The two last sections in this chapter
cover radial basis functions and Kalman
filtering in about one page and a half. Needless to say this will very likely not be enough to properly
familiarize the novice reader with these importante topics,
specially Kalman filtering, which is itself
difficult to be properly motivated and described even in books wholy dedicated to this subject.
|
|
|