We have been advised by Brainchip that Peter van der Made and his merry men at Brainchip's Perth Innovation Centre have all but finalised the design of AKD2000 which will add prediction to the AKIDA technology range. The following research article which I have just come across and extracted shows how utilising the JAST rules (ie; Spike Timing Dependant Plasticity) which Brainchip holds the rights too this predictive capability is theoretically possible. So here you have independent proof of the pudding that Brainchip will introduce in the form of AKD2000 in the not too distant future and which no doubt will have NASA very excited when you consider that it will be able to extrapolate the course of travel of a meteorite or space junk well in advance of any possible collision and cause the CubeSat or interstellar vehicle to use minimum power to slightly change course early to avoid a collision. I have extracted the introduction, the Acknowledgements and the reference to Simon Thorpe and crew and also provided the link for those who might like to read the entire academic paper:
Predictive visual motionextrapolation emerges spontaneously and without supervision at each layer of ahierarchical neural network with spike-timing-dependent plasticity
View ORCID ProfileAnthony N.Burkitt, View ORCID ProfileHinzeHogendoorn
doi: https://doi.org/10.1101/2020.08.01.232595
This article is a preprint and has not beencertified by peer review [what does this mean?]
Abstract
The fact that the transmission andprocessing of visual information in the brain takes time presents a problem forthe accurate real-time localisation of a moving object. One way this problemmight be solved is extrapolation: using an object’s past trajectory to predictits location in the present moment. Here, we investigate how a simulated in silico layeredneural network might implement such extrapolation mechanisms, and how thenecessary neural circuits might develop. We allowed an unsupervisedhierarchical network of velocity-tuned neurons to learn its connectivitythrough spike-timing dependent plasticity. We show that the temporalcontingencies between the different neural populations that are activated by anobject as it moves causes the receptive fields of higher-level neurons to shiftin the direction opposite to their preferred direction of motion. The result isthat neural populations spontaneously start to represent moving objects as beingfurther along their trajectory than where they were physically detected. Due tothe inherent delays of neural transmission, this effectively compensates for(part of) those delays by bringing the represented position of a moving objectcloser to its instantaneous position in the world. Finally, we show that thismodel accurately predicts the pattern of perceptual mislocalisation that ariseswhen human observers are required to localise a moving object relative to aflashed static object (the flash-lag effect).
Significance Statement Our ability to track and respond to rapidlychanging visual stimuli, such as a fast moving tennis ball, indicates that thebrain is capable of extrapolating the trajectory of a moving object in order topredict its current position, despite the delays that result from neuraltransmission. Here we show how the neural circuits underlying this ability canbe learned through spike-timing dependent synaptic plasticity, and that thesecircuits emerge spontaneously and without supervision. This demonstrates how theneural transmission delays can, in part, be compensated to implement theextrapolation mechanisms required to predict where a moving object is at thepresent moment.
https://www.biorxiv.org/content/10.1101/2020.08.01.232595v2.full
HH acknowledges support from theAustralian Research Council’s Discovery Projects funding scheme projectDP180102268. ANB acknowledges funding from the Australian Government, via grantAUSMURIB000001 associated with ONR MURI grant N00014-19-1-2571. We thank HamishMeffin and Stefan Bode for helpful comments on the manuscript.
References:
↵
Masquelier T, Guyonneau R, Thorpe SJ (2008) SpikeTiming Dependent Plasticity Finds the Start of Repeating Patterns in ContinuousSpike Trains, PLoS ONE 3(1):e1377.
(20min delay)