We often get asked by our clients, where the concrete advantage of AI in contrast to conventional statistics is manifested.
Our answer: One major factor, especially within the domain of deep learning (a subdomain of AI and machine learning), is the necessity of applied apriori knowledge or assumptions.
By applying conventional statistical methods, e.g. linear regression, complex feature engineering (=processing of data) often is required and apriori assumptions are made. In the case of linear regression a linear relationship between two investigated factors is assumed and the intensity of this relationship is analyzed.
In contrast to that approach, deep learning almost requires no feature engineering or apriori assumptions and knowledge. The functionality learns by itself – even the most complex relationships become transparent.
However, there is one big challenge: The amount of required data and their quality. In order to make valid predictions, huge amounts of high quality data are required, especially in healthcare.
– Christoph Kocher, SYTE