Indiveri, Giacomo, and Shih-Chii Liu. "Memory and information processing in neuromorphic systems." Proceedings of the IEEE 103.8 (2015): 1379-1397.
Interesting work on state-of-art neuromorphic architectures. Clearly, we are still in a brainstorming phase with experimental trials. Obviously, many of these architectures could have been simulated on CPU / GPUs. But practical implementation is a great source of inspiration and professional achievement. Additionally, investors' money wants to see the real thing to close the gap to the market. Fast time to market can indicate whether a promising technology will overcome the natural barriers of entry into application niches. But simulation of architectures can speed up the process and reduce implementation costs of dedicated integrated circuits.
Good Idea, simulation of architectures might speed up the process and reduce implementation costs of dedicated integrated circuits.
Simulation is the only way to convince community showing the results
In this work, we can find an interesting discussion about "Why not a simulation? ". Although we are not yet at the stage of multi-physics simulation in the suggested mapping, it is possible to incorporate the simulation of an application programming interface (API) between the alternative neuromorphic hardware and those that mimic the biological type neurons.
Ji, Yu, et al. "Optimized Mapping Spiking Neural Networks onto Network-on-Chip." Algorithms and Architectures for Parallel Processing. Springer International Publishing, 2016. 38-52.
- Badges
- Science method
- Similar topics
- Computer Science
- Computing
- Computing Methodologies