Unit - III

NEURAL NETWORK
Lecture 03
Back-Propagation (BPN)
The back propagation network is most widely used application network for data
compression. The complexity that arises due to hardware implementation makes
the existing system obsolete. This project describes that data compression
technique shuttle data. The network we have employed is the backpropagation
network.
Back Propagation Network
The back propagation training algorithm for training feed forward multilayer
network was developed by Paul Werberos and later Parker and Rummerlhart and
McClelland. The training method is known as back propagation (of errors) or the
generalized delta rule. Back propagation network uses sigmoidal activation
function.
One of the most common applications of BPN is in image processing. Some
examples with regard to this would be: identifying hand written characters,
matching a photograph of a persons face with a different photo in a database;
performing data compression on an image with minimal loss of content. Other
applications of the BPN are voice recognition, RADAR signature analysis and
stock market predication. All of these problems involve large amounts of data, and
complex relationships between the different parameters.
Data Compression
Data compression is often referred to as coding, where coding is to accomplish any
special representation of data which satisfies a given need. Information theory is
defined to be the study of efficient coding and its consequences, in the form of
speed of transmission and probability of error. Data compression may be viewed as
a branch of information theory in which the primary objective is to minimize the
amount of data to be transmitted.
A simple characterization of data compression is that it involves transforming a
string of characters in some representation (such as ASCII) into a new string (of

bits, for example) which contains the same information but whose length is as
small as possible. Data compression has important application in the areas of data
transmission and data storage. Compressing data to be stored or transmitted
reduces storage costs. When the amount of data to be transmitted is reduced, the
effect is that of increasing the capacity of the communication channel. Similarly,
compressing a file of its original size is equivalent to doubling the capacity of the
storage medium.
Data compression schemes are classified as either static or dynamic. A static
method is one in which the mapping from the set of messages to the set of code
words is fixed before transmission begins, so that a given message is represented
by the same code word every time it appers in the message ensemble like in
Huffman coding. In Huffman coding, the assignment of code words to source
messages is based on the probabilities with which the source messages appear in
the message ensemble. Messages, which appear more frequently, are represented
by short code words; messages with smaller probabilities map to longer code
words. These probabilities are determined before transmitting.
Conventional Methods of Data Compression

Huffman Coding
Shannon-fano coding
Universal codes
Arithmetic codes
Lempel-ziv Coding

Huffman Coding