T 1635/19 - Abstract scheme of machine learning

• “The application relates to adaptive remote maintenance of rolling stock with the help of machine learning.”
• Inventive step is to be examined.
• “The distinguishing features (a) to (c) express that a new rule for the identification of events/incidents from diagnostic information is created by applying supervised learning [i.e. machine learning] to the diagnostic information while using the information indicating which of the generated events/incidents have been validated and which have been discarded by human experts as ground truth information. Feature (d) then expresses that the newly generated rule is applied (to future diagnostic information).”
• “The distinguishing features (a) to (d) relate to an abstract scheme of [machine] learning a new classification rule (for classifying certain combinations of sensor values as events/incidents) on the basis of input data (the diagnostic information) and expected output data (the "ground truth information" indicating whether the events generated by the current rules are correct (by having been validated) or not (by having been discarded)). This abstract scheme is non-technical "as such". This means that the scheme is to be taken into account when assessing inventive step only to the extent that it interacts with the technical subject-matter of the claim to solve a technical problem (T 154/04, OJ EPO 2008, 46, Reasons 5, under (F)).

T 1635/19 - 

https://www.epo.org/law-practice/case-law-appeals/recent/t191635eu1.html

decision text omitted.