PROJECT TITLE :

Learning in Nonstationary Environments: A Survey

ABSTRACT:

The prevalence of mobile phones, the internet-of-things technology, and networks of sensors has led to an huge and ever increasing quantity of knowledge that are currently more commonly accessible during a streaming fashion [1]-[five]. Often, it is assumed - either implicitly or explicitly - that the method generating such a stream of knowledge is stationary, that's, the information are drawn from a mounted, albeit unknown chance distribution. In several real-world scenarios, but, such an assumption is merely not true, and the underlying method generating the info stream is characterised by an intrinsic nonstationary (or evolving or drifting) phenomenon. The nonstationarity will be due, for instance, to seasonality or periodicity effects, changes within the users' habits or preferences, hardware or software faults affecting a cyber-physical system, thermal drifts or aging effects in sensors. In such nonstationary environments, where the probabilistic properties of the info change over time, a non-adaptive model trained beneath the false stationarity assumption is sure to become obsolete in time, and perform sub-optimally at best, or fail catastrophically at worst.