High level Implementation Of Simple Self-Organzing Map (Kohonen Map)

follow by the instruction on the ai-junkie website : http://www.ai-junkie.com/ann/som/som1.html

This is the second task on ICCS internship program at TU-Chemnitz , Germany

by Theppasith Nisitsukcharoen

Third-year Computer Engineering student , Chulalongkorn University

Desktop Side

1. Windows 7 (According to serial transmission Library)
2. SFML Library - Draw Visualization

• SFML -Please Select 32-Bit - code:block set path to = C:\SFML-2.3

3. CODE::BLOCK (IDE for C++/C) - Project files formatted for code:Block

• CODE:BLOCK

Microcontroller Side

1. Source-Code here SimpleSOM-MSP430
2. MSP430F5529 Board (Texas instrument)
3. Code Composer Studio (Texas instument modded version of Eclipse)

• First, you have to clone this repo into your computer.
• You can use SOURCETREE or any git clients you want to.

   $ cd path/to/your/workspace
   $ git clone https://github.com/Tutorgaming/SimpleSOM-junkie.git

• Add "C:\SFML-2.3\bin" to WINDOWS PATH

• Then Open the project file using Code::Block [Tutor-SOM.cbp]

• Here you go :D

• Normally , the test dataset is set to "iris.data" which is located on root directory of the project file.
• There are some flag to set before compiling

/*================================
   @FLAG
==================================*/
    int                 serial_flag     = 1; //Enable Serial
    int                 finished        = 0;
    int                 classy          = 0;
    int                 num             = 0;
    int                 plotty          = 0;
    

• Also the Parameters for self organizing map network

/*================================
   @PARAMETERS
==================================*/
    unsigned const int  ROW             = 15;
    unsigned const int  COL             = 15;
    unsigned const int  MAX_ITERATION   = 250;
    double              MAX_radius      = max(ROW, COL)/2;
    const double        LEARNING_CONST  = 0.1;
    

1. Program will count lines and elements of the dataset.
2. After that, it imports dataset onto those vectors

• vector <double> data
• vector <double> real

3. Next , create the self-organizing network with randomized weight
4. User will prompt to press "ENTER" in order to start program
5. Traning Process - Bring the dataset from vector <double> data one by one to train the network until the end of dataset
6. Shuffle Dataset (vector <double> data)
7. Do Training Process until reach the iteration round ( according to MAX_ITERATION )
8. After training , Plot the dataset on the trained map ( to specify which region are the data )

• Use plotter[Row][Col]

9. Sending the trained Self-Organizing network to Microcontroller ( according to serial_flag )

• Running Microcontroller is required
• function " sentMapToMSP(); "

10. Finally, the program will ask the input test vector ( its size depends on elements of dataset ) and send it to microcontroller
11. The Result

• Shown on the console (Desktop Side)
• Blinking LED on microcontroller ( numbers depends on the class tag )