1 of 20

Lecture 4

ASSOCIATIVE MEMORY

2 of 20

Introduction

  • In computing, memory is a device or system that is used to store information for immediate use.
  • Examples: RAM, ROM,PROM, EPROM etc.
  • Data is stored in form of words and each word is accessible by an address (address addressable memory).
  • With technology advancement, we come up with the need of very high speed memory….i.e. CAM (Content Addressable Memory).
  • Content-addressable memory (CAM) is a special type of computer memory used in certain very-high-speed searching applications.
  • It is also known as associative memory or associative storage and compares input search data against a table of stored data, and returns the address of matching data.

​

​

3 of 20

Associative memories/CAM can be implemented either by using feedforward or recurrent neural networks. Here we will only see feed forward implementation.

​

4 of 20

Some Terminologies

  • Hamming Distance
  • Training Algorithms for Pattern Association
    • Hebb Rule
    • Outer products Rule

​

​

April 2007

4

5 of 20

Hamming Distance

April 2007

5

6 of 20

HEBB RULE: Used for finding weights of an associative memory neural net. The training vector pairs are denoted as s:t�

April 2007

7 of 20

HEBB RULE ALGORITHM

April 2007

7

8 of 20

Outer Products Rule: alternative method for finding weights of an associative net

April 2007

8

9 of 20

Comparing the results of Hebb Rule and Outer Product Rule

April 2007

9

10 of 20

TYPES OF ASSOCIATIVE MEMORY NETWORKS

  • Two types based on type of associations between i/p (s) and o/p (t)
    • hetero-associative (s != t): relating two different patterns (s – input, t – target).
    • auto-associative (s = t): relating parts of a pattern with other parts.

​

Definitions

In the case of an autoassociative neural net, the training input and the target output vectors are the same.

In case of a heteroassociative neural net, the training input and the target output vectors are different.

11 of 20

Architectures of associative memory:�

AUTO-ASSOCIATIVE MEMORY

HETERO-ASSOCIATIVE MEMORY

​

12 of 20

Flowchart for�training process�(same for both� types)

April 2007

12

13 of 20

Training Algorithm: AUTO-ASSOCIATIVE MEMORY�

April 2007

13

14 of 20

Testing Algorithm: AUTO-ASSOCIATIVE MEMORY�

April 2007

14

15 of 20

EXAMPLE

April 2007

15

16 of 20

April 2007

16

17 of 20

Training Algorithm: HETERO-ASSOCIATIVE MEMORY�

  • SAME AS THAT OF AUTO ASSOCIATIVE MEMORY NETWORK

April 2007

17

18 of 20

Testing Algorithm: HETERO-ASSOCIATIVE MEMORY�

April 2007

18

19 of 20

EXAMPLE

April 2007

19

20 of 20

April 2007

20