It has become clear in recent times that the use of machine learning and probabilistic modelling tools is becoming more and more necessary to deal with the large amounts of data which are now available in astronomy, and with the complex latent variable questions which are being asked. Furthermore interactions of astronomers and those from the machine learning community can bring insights into new techniques which are useful both in astronomy but also in the wider arena. Edinburgh University is funding a new initiative to push forward research on joint work between machine learning people, computer scientists and astronomers. Chris Williams and Andrew Lawrence and the principle investigators on this first project. We are pursuing a number of issues and problems where there is a significant overlap of interest. One of the early primers, just to illustrate that things can be achieved involved developing techniques for finding satellite tracks within large sky survey datasets. Other issues currently being studied include automatic categorisation of stars and galaxies, the problems of record linkage and estimating star generation histories in galaxies.