A User ' s Manual and Guide to an SDM

In this document we brieey describe the Sparse Distributed Memory model and a menu-based system which allows a user to gain hands-on experience with this model. The system is designed to aid beginners by providing stored exercises and examples. It also provides experienced users with options for connguring the memory and running their own learning experiments. The system is implemented in the *Lisp language on a Connection Machine. 1 Background A sparse distributed memory (SDM) is a model of human long-term memory developed by Pennti Kanerva 3], which ooers another approach to the problem of machine learning. The model is constructed from simple, neuron-like components which makes it ideal for implementation on a parallel computer. 1.1 The Basic Features of SDM The SDM model is based on the properties of the space of n-bit binary strings, where n is large (in the range 100 to 10,000). Both the addresses to the memory and the data stored in the memory are n-bit binary strings.