So I have a couple of research ideas. And I have a meeting with one of my professors tomorrow to talk about them, and see if we could get either of them to be big enough and worthy enough of being a Master's Thesis.

One involves obtaining a whole bunch of data to try and find some correlation between input data, neural network design, and the number of hidden nodes needed in a Back Propagation algorithm to provide optimal results.

The other is an extension of Heather Chan's work, which I learned about at the Spring Research Conference. It involves using Frequency analysis to do song identification. I figure I could expand this to include volume analysis, song speed, beat detection... and do Genre classification or artist identification, or possibly song identification as well. My current adviser suggested possibly do song prediction, attempting to learn what the next sound would be in a song, that could be interesting as well.

[This Entry]