Using data to organize large sets of analysis methods
Nick Jones (Imperial College)
Friday 8th February, 2013 15:00-16:00 Maths 204
I will talk about our efforts to create large scale libraries of both time series methods and time series data. We implemented more than 1000 different time series algorithms (with ~10 variants each) and then studied their behaviour on >30000 empirical, and model generated, signals. We used the data to organize our methods to yield a data-driven approach to organizing time series tools.
If I have time I will also discuss our work investigating mitochondrial variability which requires a variety of stochastic models and inference tasks.