Emerging mobile applications rely upon knowing a user’s location. A (geospatial) position is a low-level conception of location. A place is a high-level, user-centric conception of location that corresponds to a well-delineated set of positions. Place recognition deals with how to identify a place. Traditional place-recognition approaches (1) presuppose manual tuning of place parameters;(2) limit themselves to specific sensors; or (3) require frequent power-consuming sensor readings. We propose Platys, an adaptive, semisupervised approach for place recognition, which makes weak assumptions about place parameters, and the types and frequencies of sensor readings available. We evaluate Platys via a study of six users. A comparison with two traditional approaches indicates that Platys (without parameter tuning) performs better than traditional approaches (with optimally tuned parameters).