BACK TO PROJECT PAGE
July 27th
No entry
July 28th
No entry
July 29th
Today I finished my paper on protein analysis and SOM. I discovered that SOM is a great tool for classifying proteins. However, another neural networking method would be better for predicting secondary structure of proteins. I think that SOM will be very useful overall for classifying proteins and I think that continued research will be valuable for medicine and genetic research. CLICK HERE TO READ MY PAPER ON SOM AND PROTEIN ANALYSIS
July 30th
Today I met with Professor Valova and I showed her my program with protein recognition. She and I both agreed that some of the figure windows in MATLAB should be larger. We agreed to try to increase the performance of the map by increasing the number of epochs and increasing the map size. As of now the program runs with 300 training epochs and a map size of 10 x 10. I am going to increase the performance to 500 training steps and the map size to 15 x 15. Hopefully that allows the SOM to capture the entire data set.
July 31st
Today I increased the training steps to 500 and I think that 1000 would even be better. I ran the program with 1000 training steps and it does in fact allow the SOM to capture more of the data. I think it will be better if the map size was larger. I increased the map size to 15x15 and it works better. The only problem is that when the map size is larger the program works slower, which is not that big of a deal. The only issue with the SOM is that on some training steps the map begins to twist and this hinders it from stretching far enough to reach all of the data.
August 1st
Today I tried a third encoding. The first and second encoding worked well. This third encoding does not perform well. It basically takes the letter of the amino acid and assigns it an even number. For instance, alanine is A therefore A is 2. So the data set is more disorganized not as linear as the second encoding. I was hoping to find an encoding that would also allow the map to capture the entire data, but so far the second encoding works the best. Professor Valova wants me to present something on SOM to the incoming freshmen. We have agreed that I will present my MATLAB programs on both ART and SOM using power point. Then to catch their attention I am going to tell them a little something about how SOM is used in facial recognition. I think that will interest the students more than proteins.
August 2nd
Today I am working on finishing up my website with all of my code for SOM and the output from MATLAB. I am also trying to find some more research on face recognition. I want something that explains how SOM works with face recognition without having to go into so much technical detail of the programming. Some freshmen might know a little about programming and neural networking but for those who donít I want them to understand the concept without overwhelming them. I want to keep them awake so I want my presentation to be more visual than just me talking and talking. CLICK HERE TO SEE MY RESULTS FROM MY PROGRAM.
CONTINUE READING MY LOG ENTRIES
BACK TO LOG ENTRIES
