Download presentation

Presentation is loading. Please wait.

Published byBrian Cameron Modified over 3 years ago

1
God

2
NEURAL NETWORKS M. Alborzi, Ph. D. Petroleum University of Technology October, 2001

3
OUTLINE Neural Networks Defined Neural Networks Defined Why Neural Networks Why Neural Networks Pattern Recognition Pattern Recognition Neural Networks Application Areas Neural Networks Application Areas A Brief History of Neural Networks A Brief History of Neural Networks Training Neural Networks Training Neural Networks Advantages of Neural Networks Advantages of Neural Networks A Simple NN Package A Simple NN Package

4
Neural Networks Defined A Modeling Technique Emulating the Brain A Modeling Technique Emulating the Brain

6
Why Neural Networks! The Need to Emulate the Brain The Need to Emulate the Brain Facing Complex Problems Facing Complex Problems Limitation of Mathematics Limitation of Mathematics Limitation of Serial Computers Limitation of Serial Computers The Amazing Power of the Brain to Tackle complexities The Amazing Power of the Brain to Tackle complexities The Parallel Nature and the Network Nature Structure of the Brain The Parallel Nature and the Network Nature Structure of the Brain

7
Pattern Recognition Mathematical / Statistical Mathematical / Statistical Syntactical Syntactical Neural Networks Neural Networks

8
Neural Networks Applications in Pattern Classification and Pattern Recognition Speech recognition and speech generation Prediction of financial indices such as currency exchange rates Location of radar point sources Optimization of chemical processes Target recognition and mine detection Identification of cancerous cells Recognition of chromosomal abnormalities Detection of ventricular fibrillation Prediction of re-entry trajectories of spacecraft Automatic recognition of handwritten characters Sexing of faces Recognition of coins of different denominations Solution of optimal routing problems such as theTraveling Salesman Problem Discrimination of chaos from noise in the prediction of time series

9
A Brief History of Neural Networks 1943 McCulloch and Pitts Model 1962 Rosenblatt Perceptron 1969 Miskey and Papert Report on the Shortcomings of Perceptron 1987 Rumelhart and McClleland Breakthrough, Multilayer Perceptron (Originally from Werbos),

10
Figure 1: The biological neuron

11
Y= fh[sum( wixi)-teta] fh(x)=1 if x>0 fh(x)=0 if x<0 Figure 2: The McCulloch and Pitts model of a neuron. X1 X2 X3 OUT

12
Figure 3: A comparison between M & P model of a neuron and the biological neuron. M-P model Biological Neuron ------------------------------------------------------------ Input data x i ---------------------------Input signal Input branches------------------------Dendrites Weights w ji ----------------------------Synapses w ji x i -----------------------------------Activation Threshold L---------------------------Threshold level Output yj------------------------------Output signal Output branch------------------------Axon

13
XOR

14
Figure 4: Final connection weights: Positive reinforcing connections: Fixed k.

15
Figure 5: The input logs and the output dominant rock lithologies.

16
Figure 6: schematic diagram of the initial model

17
No.LogUnitDescription 1DT s/ft Sonic Velocity 2ROHBg/cm 3 Bulk Density 3NPHIPUNeutron Porosity 4PEFbarn/electronPhotoelectric Factor 5GRAPIGamma Ray Table 1: The input logs

18
Table 2: The output rock lithologies. No.SymbolUnitDescription 1DOLOFractionVolume of Dolomite 2LIMEFractionVolume of Limestone 3SANDFractionVolume of Sandstone 4ANHYFractionVolume of Anhydrite 5SHALFractionVolume of Shale

19
Appendix H A Sample of Log Measurements and PETROS Output for Gachsaran Well No. 6 1) Input Log Measuremwents DepthLog Measurements metresDTROHBNPHIPEFGR s/ft g/cm 3 PUbarn/electronAPI 2505.0052.7002.8201.2204.82034.100 2505.1552.8002.8001.4704.67033.600 2505.3052.7002.7901.5404.64030.400... 2667.3049.2002.7403.8704.59023.000 2667.4649.1002.7203.8804.63023.000 2667.6149.1002.7203.8804.68023.000 A Sample of Log Measurements and PETROS Output for Well No. 6 1) Input Log Measuremwents

20
DepthVolume Fractions of the Rock Constituents metresDOLOLIMESANDANHYSHAL fraction 2505.000.4200.0000.2600.2400.080 2505.150.5000.0000.3000.1200.080 2505.300.5200.0000.3000.1000.080... 2667.300.4200.5800.000 2667.460.3800.6200.000 2667.610.3600.6400.000 2) PETROS Output Volume Fractions

Similar presentations

OK

Artificial Intelligence Methods Neural Networks Lecture 1 Rakesh K. Bissoondeeal Rakesh K.

Artificial Intelligence Methods Neural Networks Lecture 1 Rakesh K. Bissoondeeal Rakesh K.

© 2017 SlidePlayer.com Inc.

All rights reserved.

Ads by Google

Ppt on stock market download Ppt on led driver A ppt on natural disasters Ppt on writing equations from tables Ppt on albert einstein biography Ppt on mapping space around us Ppt on aerobics exercise Ppt on p&g products list Ppt on school library management system Ppt on charge coupled device detector