Biologically inspired spiking neural networks are increasingly popular in the field of artificial intelligence due to their ability to solve complex problems while being power efficient. They do so by leveraging the timing of discrete spikes as main information carrier. Though, industrial applications are still lacking, partially because the question of how to encode incoming data into discrete spike events cannot be uniformly answered. In this paper, we summarise the signal encoding schemes presented in the literature and propose a uniform nomenclature to prevent the vague usage of ambiguous definitions. Therefore we survey both, the theoretical foundations as well as applications of the encoding schemes. This work provides a foundation in spiking signal encoding and gives an overview over different application-oriented implementations which utilise the schemes.

No abstract

Spiking neural networks offer the potential to drastically reduce energy consumption in edge devices. Unfortunately they are overshadowed by today's common analog neural networks, whose superior backpropagation-based learning algorithms frequently demonstrate superhuman performance on different tasks. The best accuracies in spiking networks are achieved by training analog networks and converting them. Still, during runtime many simulation time steps are needed until they converge. To improve the simulation time we evaluate two inference optimization algorithms and propose an additional method for error minimization. We assess them on Residual Networks of different sizes, up to ResNet101. The combination of all three is evaluated on a large scale with a RetinaNet on the COCO dataset. Our experiments show that all optimization algorithms combined can speed up the inference process by a factor of ten. Additionally, the accuracy loss between the original and the converted network is less than half a percent, which is the lowest on a complex dataset reported to date.

Machine learning applications are steadily increasing in performance, while also being deployed on a growing number of devices with limited energy resources. To minimize this trade-off, researchers are continually looking for more energy efficient solutions. A promising field involves the use of spiking neural networks in combination with neuromorphic hardware, significantly reducing energy consumption since energy is only consumed as information is being processed. However, as their learning algorithms lag behind conventional neural networks trained with backpropagation, not many applications can be found today. The highest levels of accuracy can be achieved by converting networks that are trained with backpropagation to spiking networks. Spiking neural networks can show nearly the same performance in fully connected and convolutional networks. The conversion of recurrent networks has been shown to be challenging. However, recent progress with transformer networks could change this. This type of network not only consists of modules that can easily be converted, but also shows the best accuracy levels for different machine learning tasks. In this work, we present a method to convert the transformer architecture to networks of spiking neurons. With only minimal conversion loss, our approach can be used for processing sequential data with very high accuracy while offering the possibility of reductions in energy consumption.

Many hand gesture recognition systems use radar to sense the motion of the hand due to its independence of lighting and its inherent privacy. As in the case of cameras, complex signal processing chains consisting of classical algorithms and neural network-base approaches are necessary to evaluate the incoming data stream. Especially on mobile devices, the reduction of the total energy consumption of the recognition system is crucial as it would lead to an increased battery life. Spiking neural networks have been shown to consume much less energy than current networks by operating event-driven and using time as the main information carrier. However, practical applications in which they are on par with classical approaches are rare. In this paper we utilize spiking neural networks to perform hand gesture recognition in radar data. We show that the temporal affinity of spiking networks and the possibility to binarize the radar-generated range-Doppler images without large loss of information introduces a promising synergy. Using simple networks consisting of 75 recurrently connected spiking neurons, we are able to reach current state-of-the-art performance on two public datasets. With this approach, gesture recognition systems can operate much more energy-efficient, making spiking neural networks viable alternatives to current solutions.