The HR project (standing for Hardy-Ramanujan, two famous pure mathematicians from the last century) started out as an investigation into how we can automated mathematical creativity, most notably concept formation and conjecture making. Three large scale systems have been built by Simon Colton, and the project is long-standing in Computational Creativity research (with first publications going back to the late 90s). These systems have been used as stand-alone AI systems in addition to being used in large combined reasoning systems, and have led to the discovery of a number of new mathematical concepts, conjectures and theorems.

The current version of the software is called HR3 and has been re-imagined from scratch as an automated programming system, fitting into a larger programme of research in the field on software writing software. In particular, HR3 forms algorithms which manipulate data in order to discover patterns in the data. However, the point of the exercise is not to highlight the patterns – this is just a welcome side-effect – but rather to create algorithms which have more application than just to the given dataset. In this way, we see it acting more as a machine learning researcher, who creates generic algorithms for data mining, than as a machine learning approach, which simply applies one or more of these generic algorithms.

A technical description of HR3 can be found in this paper:

The HR3 Discovery System: Design Decisions and Implementation Details

We have applied it to solve the Countdown Numbers game, in the sense of calculating optimal solutions to all problem instances, as described in this paper:

Countdown Numbers Game: Solved, Analysed, Extended

For further details of the HR project, see sections 1 to 6 of this page.