2021 BRN Discussion, page-5829

The following article from the Democritus University of Thrace is quite interesting. They do bearing fault analysis using the same Case Western Reserve University (CWRU) dataset used by Brainchip for this same purpose.
In it they developed a CNN algorithm called 'Attention stream net' which obtains the very high accuracy for a high number of cases but with only a small percentage of the training samples.
In other words, by only requiring a relatively small amount of samples and being able to distinguish classes with high accuracy, it is more ideal for edge (Akida) applications.



https://www.researchgate.net/publication/350296686_Extracting_spatially_global_and_local_attentive_features_for_rolling_bearing_fault_diagnosis_in_electrical_machines_using_attention_stream_networks

Note that there is no mention of spiking neural networks or neuromorphic computing, so this study doesn't seem to have used Akida.
However, IMO this indicates more of an ongoing partnership between them and Brainchip.

Pure speculation, DYOR