Teaching Demos

Click the links to download the demonstration software.

• SOM Demo: Demonstrates the Self-Organizing Map machine learning algorithm. The application generates random data and then users can step through the algorithm, viewing the intermediate results. See below for instructions.
  Design and Implementation: Michael Rice, Wesleyan University.

  To install the software, first download the above zip file to your computer (Windows machines only) and extract the installation folder. Then run SOM->Package->setup.exe to start the installation wizard.

• K-Means Demo: Demonstrates the K-Means clustering algorithm. The application generates random data and then users can step through the algorithm, viewing the intermediate results. See below for instructions.
  Design and Implementation: Michael Rice, Wesleyan University.

  To install the software, first download the above zip file to your computer (Windows machines only) and extract the installation folder. Then run KMeansDemo->Package->setup.exe to start the installation wizard.

---

Teaching Demo Instructions

SOM Demo

The RGB values (e.g. (100, 255, 255)) in each square in the grid represent a model vector of theoretical expression values in three microarrays.

The model vectors in the grid are updated in each iteration based on fixed expression vectors for genes. (Note that the gene expression vectors are not shown in the demo.) Using the SOM algorithm, only the model vector closest to each gene vector, and the grid squares within the radius distance, are updated.

To run the demo:

1. Choose the following setting:
   -- size of grid (e.g. 8 x 8)
   -- number of data points (e.g. 8 genes)
   -- number of iterations performed for each update (e.g. 1)
   -- radius of cells updated (e.g. 5)
   
2. Click "Create Grid" (see * below)
   
3. (a) - To update the grid based on one data point (gene) at a time, press "Update One"
   (b) - To update the grid based on all data points (genes) at once, press "Update All"
   
4. Continue updating until the grid values stabilize (Note the number of iterations and pattern of colors representing RGB model vectors of gene expression)
   
5. Once the model vectors in the grid have stabilized, each gene is assigned to the closest model vector (grid square), thereby defining clusters
   
6. You may explore the effects of varying parameters; for example, try using 100 data points (genes)
   

[* The "Create Grid" button assigns random values to each model vector in the grid and initializes the RGB value for each data point (gene) to the same value if the number of data points (genes) is not changed. The "Create Grid (R)" button initializes the RGB values for each data point (gene) to random values. ]

K-Means Demo

1. Click add random data (should put specified number of points in first panel).
2. At this point, the three option buttons below the button you just clicked should have the "add means" selected. If not, select that option and then click at a few points in the first panel to add the initial positions of the green means.
3. Click "distribute data" to see the clusters visually in the first panel.
4. Click in the second panel and then click "update means" to display the data points and positions of new means in the second panel.
5. Repeat 3 and 4 until convergence.