Whilst large scale Artificial Intelligence models contain to gain an increasing amount of attention, for many scenarios modelling tabular data, more traditional statistical machine learning approaches will give models that are more transparent, interpretable and robust.

Although many approaches in this area are well tried and tested, research continues to develop improvements to existing methodologies.  I was honoured to be recognised as having made a small contribution to this paper describing UniLasso, the latest development of the Lasso methodology  developed by Rob Tibshirani and Trevor Hastie, and to be invited to write a  Commentary on the paper.