{[ promptMessage ]}

Bookmark it

{[ promptMessage ]}

analog circuits(network systems)..

analog circuits(network systems).. - By...

Info iconThis preview shows pages 1–3. Sign up to view the full content.

View Full Document Right Arrow Icon
Analog Circuits for Self-Organizing Neural  Networks Based on Mutual Information By: R. Nitisha (CSE IInd Year) [email protected] S. Kalyani (CSE IInd Year) [email protected] KITS WARANGAL Abstract    A  new   self-organizing   neural   network   concept   based   on   mutual  information is described in this paper. Comparing to conventional neural  network structures, this organization greatly reduces the interconnections in  neural network by using local interconnection based on statistical analysis,  and eliminates the need to store large number of synaptic weights. The  network is characterized by evolvable hardware structure and adjustable  threshold values based on selection criteria, which use mutual information.  Secondly,   A   mix-signal   implementation   scheme   is   proposed   for   this  organization   in   order   to   achieve   the   best   performance.   The   digital  implementation is used for the evolvable structure of the network for the  better ability to be reconfigured. Analog implementation is used for the  entropy-based evaluator (EBE), which is used for statistical analysis and  mutual information evaluation, in order to achieve smaller area and faster 
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full Document Right Arrow Icon
on-chip learning process. Either on-chip analog memory or off-chip digital  memory can be used to store the threshold values of the neurons and  organization of resulting interconnection of neurons. Finally, circuits used  for the analog implementation of the EBE are presented; the simulation  results of the circuits are shown and discussed. 1 INTRODUCTION An artificial neural network (ANN) is a massively parallel distributed  processor made up of simple processing units, which has a natural ability  for storing experimental knowledge and making it available for use.[1] In  general, there are many different classes of network architectures. All of  them   require   extensive   interconnections   between   source   nodes   and  neurons,   and/or   between   neurons.   The   knowledge   acquired   through   a  learning process is stored as synaptic weights for every connection. Since  the number of interconnections and synaptic weights increase quickly as  the number of input nodes and neurons increases, it is difficult to build a  neural network with large number of neurons, even with the currently most  advanced VLSI fabrication technology.  To   overcome   this   difficulty,   a   new   reconfigurable   artificial   neural  network organization using learning algorithm based on maximum mutual  information principle (MMIP) and statistical analysis was developed for data  classification application.[2] This organization also uses multi-layer feed 
Background image of page 2
Image of page 3
This is the end of the preview. Sign up to access the rest of the document.

{[ snackBarMessage ]}

Page1 / 14

analog circuits(network systems).. - By...

This preview shows document pages 1 - 3. Sign up to view the full document.

View Full Document Right Arrow Icon bookmark
Ask a homework question - tutors are online