In the past few months, in my spare time whenever I get a chance, I’ve been working on a side project in work. It’s a research project amongst a group of people with a cool idea. The idea is to apply neural networks to identifying diseases, populations and traits via machine learning/supervised learning. I’m not entirely sure about the application of the technology as to how new or revolutionary it is, but it seems sensible that a machine is going to be very good at being taught how to recognise patterns so why not do it.
I’m not directly responsible for the code for this paper, but I did go back and refactor most of the code in an attempt to make it a little presentable to the world and also much faster as a result. I hope that the source code will eventually see the light of day.
Note, I am not the primary author of this presentation/paper, I just have an interest in the idea, code and enjoy learning and playing with this research.
There will be a presentation at EMGM, I will be curious as to what people think. Is it a crazy good or a crazy bad idea? The data that is being generated is apparently good, though personally I would like to see more analysis being done with additional ANN’s or SVM’s to verify that it really is working as expected.
I wonder if I can convince a CS or a Maths summer student to re-implement the algorithm with a different ANN or SVM library. There are a rake of ideas that I have that I wouldn’t mind seeing implemented. I just need more time!