And the proposed uses just keep lining up:
“3 CONCLUSION
We present a framework for symbolic computing with spiking neurons based on KGs and graph embedding algorithms. Compared to previous approaches based on semantic pointers [8], our method allows the learning of semantically meaningful, low-dimensional and purely spike-based representations of graph elements. Due to the differentiability of our approach, such semantic representations can be trained end-to-end in unison with other differentiable models or architectures, such as multi-layer or recurrent SNNs, for specific use-cases, as demonstrated with our spiking GNN. In addition, by changing the scoring function (Eq. 2), alternative spike-based embedding schemes can be devised in future work that leverage gradient-based optimization in SNNs – similarly to how a variety of embedding schemes exist in the graph embedding literature [27].
The combination of symbolic data and spike-based computing is of particular interest for emerging neuromorphic technologies [11], as it bears the potential of opening up new data formats and applications. Thus, we are convinced that our work constitutes an important step towards not only enabling, but also exploring neuro-symbolic reasoning with spiking systems.
ACKNOWLEDGMENTS
We thank Serghei Mogoreanu, Alexander Hadjiivanov and Gabriele Meoni for helpful discussions. We further thank our colleagues at the Semantics and Reasoning Research Group, the Siemens AI Lab and ESA’s Advanced Concepts Team for their support. This work was partially funded by the Federal Ministry for Economic Affairs and Energy of Germany (IIP-Ecosphere Project) and by the German Federal Ministry for Education and Research (“MLWin”, grant 01IS18050). DD acknowledges support through the European Space Agency fellowship programme.”
What was it about AKIDA out performing GPUs???
My opinion only DYOR
FF
AKIDA BALLISTA
