With evolutionary computing techniques, you don’t need to solve technical problems, merely create a system for evaluating solutions. I recently wrote a paper on using evolutionary computing methods to detect diseases such as COVID-19 using evolutionary methods. This chart shows different trials of evolving solutions to handwriting recognition.
Evolutionary computing, means encoding computer programs into “DNA” of binary strings. Each string represents a solution to the problem being solved. Solutions are tested, selected and combined to form new candidate solutions. The incredible nature of evolutionary search makes is suitable to challenging problems for which analytical solutions are unknown or impossible. This type of search has an uncanny way of searching n^3 terms for n evaluations, due to the schema theorem.
My hopes to apply this technique to COVID-19 ran out because I lacked sufficient funding to continue renting expensive parallel GPU configurations. However, the work I did to demonstrate speedup and parallelize evaluations in Keras are still pretty useful. If I wasn’t busy working on my thesis, I would like to continue this project. Perhaps someone else can benefit from the Github of source code.