Incremental Learning is a key feature when dealing with tiny and embedded devices. Many have claimed that only inference can be performed on embedded devices. At Ekkono, we have built a library that can train and run machine learning algorithms on the devices. And not only on big machines, we will do it efficiently on tiny microcontrollers. Our SDK brings incremental learning directly to your device, with no need for an internet connection.
Incremental learning uses streaming data during training. We continuously learn over time as more data and insights are being gathered. The model is being built and updated as the data arrives, and predictions are made at any point in time.
Traditional machine learning algorithms in an IoT setting requires collecting data from many different devices. This data is then used to build a generic algorithm that is representative of all devices. To collect the data, extra processing on the device is needed to send the data to the cloud or a specific machine. When there is a change in the environment, the old data will no longer be valid. This happens frequently in real life settings. For example, when one part of a machine is replaced with another, it’s behavior is often changed. When this happens, data needs to be gathered again and a new model needs to be built. This results in energy and power consumption costs, as well as extra work.
Incremental learning builds a personalized model for each device. It is built without the need to send any data to the cloud. Since the algorithm runs directly on the device, the model is adaptive to the changing environment in real-time.
The algorithm might be less general, since it is learning from only that device. On the other hand, it is also more accurate since not all devices operate under the same conditions.
See all research areas.