Presentation is loading. Please wait.

Presentation is loading. Please wait.

1 Adaptive Resonance Theory. 2 INTRODUCTION Adaptive resonance theory (ART) was developed by Carpenter and Grossberg[1987a] ART refers to the class of.

Similar presentations


Presentation on theme: "1 Adaptive Resonance Theory. 2 INTRODUCTION Adaptive resonance theory (ART) was developed by Carpenter and Grossberg[1987a] ART refers to the class of."— Presentation transcript:

1 1 Adaptive Resonance Theory

2 2 INTRODUCTION Adaptive resonance theory (ART) was developed by Carpenter and Grossberg[1987a] ART refers to the class of self organizing neural architecture that clusters the pattern space and produce appropriate weight vector ART1 – Clustering binary vectors ART2 – Accepts the continuous – valued vector Unsupervised learning ART nets are designed to be both stable an plastic ART network is a vector classifier Changes in the activation of units and weights are governed by the differential equations

3 3 Contd…. Resonance period: Activations are assumed to be change much more rapidly than the weights Once the acceptable cluster unit is selected for learning,the weights may be maintained over an extended period During that period only weight changes should be done This period is called ‘resonance period’

4 Basic Architecture 4

5 5 ART involves three group of neurons Input processing(F1 layer)  Input portion(F1(a))  Interface portion(F1(b)  Comparing the similarity of the input signal to the weight vector of the cluster unit which is selected for learning  F1(b) is connected to F2 through bottom up weights bij  F2 is connected to FI(b) through top down weight tij  Cluster unit(F2 layer)  competitive layer  Cluster unit with largest net input is selected Basic Architecture

6 6 Contd…. Activation of all other unit is set to zero Reset Mechanism Depending upon the similarity between the top-down weight and the input vector, the cluster unit may or may not be allowed to learn the pattern If the cluster unit is not allowed to learn,it is inhibited and a new cluster is selected as the candidate

7 Basic operation  Learning trial- Before the pattern is presented Activation of all units should be zero F2 units are made inactive once a pattern is given to the network,it continuously send the input signal  Controlling the degree of similarity controlled by the vigilance parameter  Reset mechanism states Function is to control the state of each node in F2 layer 7

8  Active- F2 unit is on activation is d=1  Inactive- F2 is off, activation=0 But available to participate in the next competition  Inhibit –F2 is off, activation =0, prevented from participation in further competition 8

9 Learning in ART Fast learning Used in ART1 network Input is binary Weight update occur more rapidly during resonance period Weight reaches the equilibrium on each trial Weight associated with cluster units are stabilized 9

10 Slow learning Used in ART2 network Weight changes during resonance period occur slowly Weight does not reaches the equilibrium in each trial Many more learning pattern is required Network will not be stablilzed 10

11 Basic training step 11

12 Contd….. 12

13 ART1 ART1 has Computational unit Supplemental unit 13

14 ART1 14 consist of Computational unit & supplemantal unit

15 Supplemental unit 15

16 parameter used inAlgorithm 16

17 Training algorithm 17 Step 0 : initialize parameters : initialize weights :

18 18 Adaptive Resonance Theory NN 18 Step 1: While stopping condition is false do Steps 2-13 Step 2: For each training input. do steps 3-12 Step 3: Set activations of all F2 units to zero. Set activations of F1(a) units to input vector s. Step 4: Compute the norm of s: Step 5: Send input signal from F1(a) to the F1(b) layer ART1 Algorithm (cont.)

19 19 Adaptive Resonance Theory NN 19 Step 6: For each F2 node that is not inhibited: if. then Step 7: While reset is true. do Steps 8-11. Step 8: find J such that y J ≥y j for all nodes j. If y J then all nodes are inhibited and this pattern cannot be clustered. Step 9: Recompute activation x of F1(b) x i = s i t Ji ART1 Algorithm (cont.)

20 20 Adaptive Resonance Theory NN 20 Step 10: Compute the norm of vector x: Step 11: Test for reset: if then y J =-1 (inhibit node J)(and continue executing step 7 again) Ifthen proceed to step 12. ART1 Algorithm (cont.)

21 21 Adaptive Resonance Theory NN 21 Step 12: Update the weight for node J (fast learning) Step 13: Test for stopping condition. ART1 Algorithm (cont.)

22 22

23 23

24 24

25 25

26 26

27 27

28 28

29 29

30 30

31 ` THANK YOU 31


Download ppt "1 Adaptive Resonance Theory. 2 INTRODUCTION Adaptive resonance theory (ART) was developed by Carpenter and Grossberg[1987a] ART refers to the class of."

Similar presentations


Ads by Google