I have brought with me one of the SpiNNaker2 single-chip test boards, and I am planning to show a simple demo to showcase the hybrid properties in SpiNNaker2. The example is modelled and trained in Python using TensorFlow, before it is deployed to SpiNNaker2.
SpiNNaker2 supports the native execution of DNNs and SNNs under the same application. Such setup is useful for hybrid models that can leverage native (non-spiking) DNN-based feature extraction (e.g., conv layers) as input layers of a spiking neural network, or new architectures leveraging properties from both worlds.
In order to meet the biosafety rules, I will use a TV placed in the Terrace at 16:00 tomorrow.
I have brought with me an early SpiNNaker2 test board, and
|Wed, 11.05.2022||16:00 - 17:00||Terrace|