Neural Networks and Applications

1,589 views

Published on

Published in: Science, Technology
0 Comments
6 Likes
Statistics
Notes
  • Be the first to comment

No Downloads
Views
Total views
1,589
On SlideShare
0
From Embeds
0
Number of Embeds
9
Actions
Shares
0
Downloads
0
Comments
0
Likes
6
Embeds 0
No embeds

No notes for slide
  • \n
  • \n
  • \n
  • \n
  • \n
  • \n
  • \n
  • \n
  • \n
  • \n
  • \n
  • \n
  • \n
  • \n
  • \n
  • \n
  • \n
  • \n
  • \n
  • Neural Networks and Applications

    1. 1. Neural Networks & Applications Oleksiy Varfolomiyev NJIT, 2011
    2. 2. Why Neural Networks? State-of-the-art solutions for thediversity of applications
    3. 3. What do Neural Networks do?• Clustering, classification, categorization• Function approximation• Prediction• Optimization• Associative memory• Control
    4. 4. What Applications? Manufacturing process controlCredit application Manipulator controllers evaluation Image/data compression Autopilot Breast cancer cell analysis Vehicle scheduling Special Effects
    5. 5. What Industries? Manufacturing RoboticsBanking Telecommunications Aerospace Medical Transportation Entertainment
    6. 6. Simple NeuronSimple Neuron Model !"# $%&()#&*(+ ,%-+-&. ,+/01 $/2 &#%2(+ &#*3/214 -4 *"# 4-&. &#%2/&7 4%0" (4 *"-4 #8()6+#9 !"#2# (2# *"2## -4*-&0* $%&0*-/&(+ /6#2(*-/&4 *"(* *(1# 6+(0# -& • &#%2/&9 input*"# 40(+(2 -&6%* ! -4 )%+*-6+-# ,; *"# 40(+(2 3#-. p - :-24*7 • w - weight *"# 62/%0* "!7 (.(-& ( 40(+(29 <#0/&7 *"# 3#-."*# -&6%* "! - *"# 40(+(2 ,-(4 # */ $/2) *"# &#* -&6%* $9 =>& *"-4 0(4#7 ;/% 0(& ? • b - bias (4 4"-$*-&. *"# $%&0*-/& % */ *"# +#$* ,; (& ()/%&* #9 !"# ,-(4 -4 • n - net input ( 3#-."*7 #80#6* *"(* -* "(4 ( 0/&4*(&* -&6%* /$ @9A :-&(++;7 *"# & 6(44# *"2/%." *"# *2(&4$#2 $%&0*-/& %7 3"-0" 62/%0#4 *"# 40(+( • !"# f - transfer */ *"#4# *"2## 62/0#44#4 (2#B *"# 3#-."* $%&0*- &()#4 .-?#& function • a - output -&6%* $%&0*-/& (& *"# *2(&4$#2 $%&0*-/&9 :/2 )(&; *;6#4 /$ &#%2(+ &#*3/2147 *"# 3#-."* $%&0*-/& -4 ( 62/
    7. 7. Neuron with Vector Input
    8. 8. One Layer of Neurons
    9. 9. Multiple Layers of Neurons
    10. 10. Neural networks Static DynamicThe output is calculated The output depends also ondirectly form the input the previous inputs, outputs, through feedforward or states of the network connections
    11. 11. Neural networks Static DynamicThe output is calculated The output depends also ondirectly form the input the previous inputs, outputs, through feedforward or states of the network connections
    12. 12. Applications ofDynamic Networks • Financial Markets • Control Systems • Fault Detection • Speech recognition • Filtering
    13. 13. The work flow for the NN design process1. Collect Data2. Create the network3. Configure the network4. Initialize the weights and biases5. Train the network6. Validate the network7. Use the network
    14. 14. Train the NetworkTuning weights and biases of the NN tooptimize NN performance function,e.g. Mean Square Error N N 1 2 1 2 F = (ei ) = (ti − ai ) N i=1 N i=1
    15. 15. Optimization methods• Use GRADIENT of the network performance w.r.t. the network weights• Use JACOBIAN of the network errors w.r.t. the network weights
    16. 16. What tools?MATLAB Neural Network Toolbox Simulink
    17. 17. Control Systems ExampleNeural Predictive Control for the Aiming and Stabilizing System
    18. 18. NN Training Error using Levenberg Marquardt algorithm
    19. 19. Control System Works Out the Harmonic Input
    20. 20. g{tÇ~lÉâ 4

    ×