Artificial Neural Networks

1
Artificial Neural Networks

• URI BME
• Aleksey Gladkov

2
Introduction

• An artificial neural network is a mathematical
  or computational model that approximates the
  structure or function of biological neural
  networks.
• (Pictured Alvin)

3
Artificial Neurons

• Artificial Neurons are modeled using a
  function which responds to various weighted
  inputs, and is capable of adjusting the weights
  over time as it learns.

4
Uses

• Artificial neural networks can be used to
  model complex relationships between many
  variables, as well as being able to spot patterns
  in a large quantities of data.

5
Drawbacks

• A major problem with artificial neural
  networks is the amount of work that must be put
  into the learning step of development. Another
  issue is the amount of processing and storage
  required to maintain such a network with the
  currently available technology.

6
Applications

• -Facial Recognition
• -Manufacturing Process and Quality Control
• -Handwriting and Speech Recognition
• -Spam Filtering
• -Gene Recognition
• -Many Kinds of Forecasting
• -Physical System Modeling

7
Memresistors

• A memresistor is a two terminal device which
  changes its resistive properties depending on the
  direction of current passing through the device.
  The most important feature of this substance is
  the ability to retain its resistive properties
  even when there is no current present.
• Memristor theory was formulated and named by
  Leon Chua in a 1961 paper.
• In 2008 HP Labs announced the development of a
  switching memristor based on a thin film of
  titanium dioxide.

8
What This Means

• Solid-state memristors can be combined into
  devices called crossbar latches, which could
  replace transistors in future computers, taking
  up a much smaller area.
• HP prototyped a crossbar latch memory using
  the devices that can fit 100 gigabits in a square
  centimeter.

9
Fuzzy Logic

• Unlike Binary Logic, which has exact values
  for TRUE and FALSE (0,1 respectively), Fuzzy
  Logic has been extended to handle the concept of
  partial truth, where the truth value may range
  between completely true and completely false.

10
What This Means

• Fuzzy logic is a lot more realistic than
  binary logic, so it can make the artificial
  neuron algorithm more realistic and lifelike.

11
Current Research

• At this time, researchers are attempting to
  evaluate the effects of neuromodulators on
  natural neural networks in order to better
  understand their functions, which will,
  hopefully, give us some insight on better
  simulating them with artificial ones.

12
The Future of Artificial Neural Networks

• With further development of memresistor
  technology and miniaturization, and liberal
  application of fuzzy logic, artificial neural
  networks can accurately mimic the functions of
  natural ones. After that is only a matter of time
  before computers that learn as the go.

13
The End?
14
References

• Alyuda. 24 Sep. 2011. lthttp//www.alyuda.com/produ
  cts/forecaster/neural-network-applications.htmgt.
• Gershenson, Carlos. "Artificial Neural Networks
  for Beginners." Cornell University. 24 Sep.
  2011.lthttp//arxiv.org/ftp/cs/papers/0308/0308031.
  pdfgt.  
• Miller, Michael J. "Memristors A Flash
  Competitor that Works Like Brain Synapses." 2010.
  Forward Thinking. 26 Sep. 2011.
  lthttp//forwardthinking.pcmag.com/chips/282616-mem
  ristors-a-flash-competitor-that-works-like-brain-s
  ynapsesfbidx6JCeArNPkcgt.
• Regine. "BRAINWAVE Common Senses." 26 Sep. 2011.
  lthttp//www.we-make-money-not-art.com/archives/200
  8/03/brainwave-common-senses.phpgt.
• Stergiou, Christos and Dimitrios Siganos .
  "NEURAL NETWORKS ." Imperial College London. 24
  Sep. 2011.lthttp//www.doc.ic.ac.uk/nd/surprise_96
  /journal/vol4/cs11/report.htmlgt.
• Wikipedia. 24 Sep. 2011. lthttp//en.wikipedia.org/
  wiki/Artificial_neural_networkgt.
• All Images Courtesy of Google Images