High-performance software implementation of event-based deep learning models
Theory of Neural Systems
Description
BACKGROUND
Event-based neural network models such as biologically inspired spiking neural networks [1] have the ability to scale well [2] while being extremely energy efficient on neuromorphic hardware [5]. While many efficient implementations exist [3,4], few support learning algorithms common in the deep learning community such as backpropagation.
YOUR TASK
Create a high-performance implementation of event-based models in an appropriate language such as Rust/Elixir or using CUDA. These implementations will also support standard learning algorithms from deep learning.
Required Skills
- Good Knowledge of Rust/Elixir/C++ and/or CUDA
- Knowledge of concurrent programming concepts
- Basic Knowledge in deep learning
References
[1] Pfeiffer, Michael, and Thomas Pfeil. "Deep learning with spiking neurons: opportunities and challenges." Frontiers in neuroscience (2018): 774. (link)
[2] Jordan, J., Ippen, T., Helias, M., Kitayama, I., Sato, M., Igarashi, J., Diesmann, M. and Kunkel, S., 2018. Extremely scalable spiking neuronal network simulation code: from laptops to exascale computers. Frontiers in neuroinformatics, p.2. (link)
[3] NEST Simulator
[4] GeNN, a GPU-enhanced Neuronal Network simulation