Which Came First? The AI or the Data?
Problem Statement:
True embedded AI starts learning as soon as a device is turned on. The device perceives the world through raw data captured by a sensor. The AI must learn automatically what part of the raw data is informative and should be used for learning. Embedded devices are often resource constraint in computation, memory, and energy. Therefore, a self-learning scheme must be energy efficient. How can efficiency be increased when an embedded device starts learning as soon as it is powered on?
Task:
The goal is to design, implement, and evaluate an experimental sensor device that learns from the data it collects.
The experiment that must be executed should evaluate the amount of energy that can be saved using a self-learning scheme.
Work:
20% Theory, 60% Simulations, 20%Writing
Contact:
Jacob Kamminga, j.w.kamminga@utwente.nl, room ZI5011