diff --git a/src/lab/exp10/Illustration.html b/src/lab/exp10/Illustration.html index 000cd59e8..0022410aa 100644 --- a/src/lab/exp10/Illustration.html +++ b/src/lab/exp10/Illustration.html @@ -105,10 +105,10 @@

Weighted matching problem


Example:

Consider N=4 points as shown in Figure 1 (a). The distances between each pair of points is given in Figure 1 (b).

- +
- - + + @@ -116,11 +116,11 @@

Example:

Figure 1 (a)

The possible pairing of points are given in Figures 2 (a) to 2 (c).

- +
- - - + + + diff --git a/src/lab/exp8/Illustration.html b/src/lab/exp8/Illustration.html index c90c6ced7..13351f9fe 100644 --- a/src/lab/exp8/Illustration.html +++ b/src/lab/exp8/Illustration.html @@ -117,10 +117,10 @@

Application of SOM to travelling salesman problem

For 100 cities and a SOM with 1000 neurons in the output layer

-
Figure 2 (a): L=7.2
+ +

-
-
Figure 1: Kohonen's self-organization feature map for TSP for 100 cities and 1000 units in the output layer.
+

Effect of varying the number of units in output layer of SOM


For the travelling salesman problem of 50 cities, we consider @@ -132,11 +132,9 @@

Effect of varying the number of units in output layer of SOM

visited, which indicates the suboptimal nature of the algorithm.

For 50 cities

- - + +

-
-
Figure 2: Illustration of sub-optimal nature of the algorithm
diff --git a/src/lab/exp8/Observations.html b/src/lab/exp8/Observations.html index 218e76f35..eb9a5c58c 100644 --- a/src/lab/exp8/Observations.html +++ b/src/lab/exp8/Observations.html @@ -120,10 +120,7 @@

Solution of optimization problems

- - +

  • Observe the output of the network for different number of cities. Start from a small number of cities (such as 10), and go up diff --git a/src/lab/exp8/Tutorial.html b/src/lab/exp8/Tutorial.html index e7a62d1d9..cdec4ad15 100644 --- a/src/lab/exp8/Tutorial.html +++ b/src/lab/exp8/Tutorial.html @@ -127,10 +127,10 @@

    Architecture of SOM

    the region in the output whose neighbourhood units have similar properties.

    - + +

    -
    -
    Figure 1: Architecture of SOM
    +

    Algorithm for learning

    The training of SOM is based on the principle of competitive learning (Refer experiment 7, Artificial Neural Networks virtual labs).