About Comp-Engine

Our site has two subsites devoted to time series analysis (signal processing) and network analysis respectively. The site has been seeded with thousands of data analysis methods and tens of thousands of data sets.


We want comp-engine to be a place:
1. for the exchange of both data and data analysis methods
2. where our methods and data can be compared (the comp in the title)

Our Motivation:

Helping tie together the sciences: All parts of the sciences generate data and develop methods for probing that data (from biology to statistics to earth sciences). Given the vigorous rate at which new data analysis methods are produced it seems reasonable to compare, not just a few methods to each other, but each method to every other method. This helps expose both similarities and novelties. Ideally we would do this at a formal level but the growth rate of new methods makes this impractical. Similarly one might want to compare data. Though one might expect different disciplines to generate different types of data, this is, in practice, often not the case: it can be helpful to work out whether data one has obtained has close matches from other disciplinary contexts. If one has examples of simulated data in the database one might also be able to match experimental to simulated data and so identify candidate models or mechanisms that might have generated our data.

A powerful (complementary) analysis of particular problems: Aside from being able to put new methods, or new data, in the context of pre-existing knowledge we might also seek to investigate the structure in a given collection of (possibly tagged-up) data objects. For example one might like to understand how (and why) EEG’s from healthy and eplieptic patients differ. A large collection of methods of the kind we present (and seek to grow) allows one (with appropriate statistical control) to perform a very thorough search for which characteristic of the data might account for differences between data types. If one can identify a set of structural characteristics of the data which are particularly explanatory then this can also help in attempts at identifying suitable mathematical models (e.g. through ABC).

You can read about our work on this topic via the two individual sites: