28 Feb Solving the memory bottleneck problem of TinyML Shraddha Goled computational overhead MIT and IBM have jointly introduced MCUNetV2, a deep learning architecture that can adjust memory bandwidth to the limits of the microcontrollers.