Stumpf´s talk highlights (Session 2)
With a talked entitled “Time and condition dependent biological networks” Stumpf discussed how to deal with models, how to make them better and how to decide if our models are acceptable. Using his own words,” biological, as well as other natural, technical and social systems, are highly structured. And these structures can change over time and between conditions.” Therefore, looking at structures that are fixed in time is essentially wrong. According to Stumpf, these changes have to be taken into account whenever possible. He claimed that we actually have the statistical tools for doing so, and that the computational efforts to learn both the structure and change of these varying networks is very modest. Stumpf ended his talk by saying that “fit of a model to data is not a reliable guide as to whether a model is good or not. Model construction has to be based on sound statistics and domain expertise”.