Rijswijk, South Holland, Netherlands
I am currently Neuromorphic Architect at Innatera. This role involves assisting in the creation of architecture specifications of future chip generations. To do this successfully, continuous and clear communication has to be enabled between application engineers, software engineers and the hardware designers. As a Research Associate at the University of Manchester, I was responsible for implementing a large scale and biophysically representative model of a Cerebellum on SpiNNaker in collaboration with partners from the University of Pavia (Italy). My PhD thesis focused on enabling the SpiNNaker neuromorphic computation platform to simulate synaptogenesis, or, in other words, structural plasticity. This is one of the mechanisms through which mammalian brains learn which relies neurons altering their connections or rewiring.
Cerebellum simulations on digital neuromorphic hardware
I help support efforts by other members of the Advanced Processor Technologies research group at the University of Manchester in converting Artificial Neural Networks into Spiking Neural Networks to be run on the SpiNNaker platform. Independently, I am responsible for developing and testing an ANN online sparsification method (DEEP R) involving the training of very sparse networks (down to 1% of original connectivity) by continuously rewiring connections. This approach is biologically inspired and of much interest to me (I wrote my thesis on this exact topic, but applied to spiking networks simulated on SpiNNaker). I use Tensorflow and Keras to achieve this, making my methods freely available and easily usable by the community. The training and inference is done on NVIDIA Tesla V100, thanks to the support of the Computational Shared Facility 3 (CSF3) maintained by IT Services. The project is funded by Huawei.
I am working on enabling the SpiNNaker neuromorphic computation platform to simulate adult neurogenesis and synaptogenesis, or, in other words, structural plasticity. This is an additional mechanism through which mammalian brains learn which relies on the "birth" and "death" of neurons in specific areas of the brain and the connections they create with pre-existing ones.
I have created a spiking neural network based robot control system using the Nengo simulator running on the SpiNNaker platform.
Worked with the Advanced Processor Technologies Research Group on the SpiNNaker Project which is the Neuromorphic Computing Platform for the Human Brain Project