I will discuss machine learning as it is applied to healthcare. The particular focus of my talk is on the limitations of pure machine learning approaches when it comes to biological systems. I will compare this with pure mathematical model-based approaches, which on the one hand appear more suitable but ultimately have their own attendant set of problems. I will finish up with hybrid approaches.
Artificial Intelligence, Machine Learning and Big Data are buzzwords which have been receiving increasing attention in the past three years. The healthcare space appears ripe for digitalisation, but so far most attempts to bring data methods to this space have been disappointing. I will discuss the deeper reasons behind this situation.
I distinguish between pure machine learing approaches and mathematical modelling, to explain the relative strengths and weaknesses of each approach. Machine learning, in particular, has some important deficits when it comes to handling biological systems. But neither approach is perfect. I will finish by introducing a new type of model, a hybrid of both approaches, which holds out the promise of allowing effective application of data-driven methods whilst fitting-in with the limitations of the healthcare domain.