On edge machine learning we are constantly dealing with changing environments. As an example, the temperature will change with the seasons, as well as humidity. Other examples of a changing environment could be small changes in the way a machine, such as a gas turbine, is operating.
Time dependent data often experience variations on trends. Ekkono’s change detection can detect those small variations. By using our change detector, the model can either adapt to the new trends or signal an alarm when a change occurs. We have implemented our change detector for different customers in many different industries. In doing so, we have discovered that detecting a malfunction or a change in the behaviour of the device ahead of time will sav both time and money.
At Ekkono, we have built a change detector that can monitor data distributions in real-time. When a change is happening, we will let you know! The uniqueness of the change detector is that it can be calibrated and run both incrementally in a microcontroller and batch-trained on a standard server. We provide two types of change detectors. The first type is one dimensional and runs directly over the sensor value. The second type is multidimensional. This means that it runs on top of a machine learning model, being able to detect concept drift on multivariate data.
See all research areas.