Skip to content


To achieve the CONVOLVE objectives, a 36-month work plan has been carefully designed based on seven strongly interrelated technical and two non-technical work packages (WPs).

CONVOLVE Work Packages overview including their dependencies.

WP1: Use case and application requirements, metrics, and baseline

WP2: Self-configurable modular ULP accelerator

WP3: Composable Real-Time and Hardware Security

WP4: Algorithmic principles for ultra-low power neural network (NN) processing

focuses on new algorithms and models for ULP Neural Networks (NN). This is done by examining leading-edge strategies, such as dynamic NNs and new online learning techniques, applicable for the edge. This is applied to artificial deep networks (ANNs) and to the less well understood, but more energy–efficient spiking networks (SNNs).

WP5: Transparent and compositional programming flow

WP6: Compositional architecture DSE and SoC generation

WP7: Application mapping, benchmarking, and integrated demos

WP8: Dissemination, Communication & Exploitation of results

WP7: Application mapping, benchmarking, and integrated demos