BayesianSpikeFusion: accelerating spiking neural network inference via Bayesian fusion of early prediction.

Spiking neural networks (SNNs) have garnered significant attention due to their notable energy efficiency. However, conventional SNNs rely on spike firing frequency to encode information, necessitating a fixed sampling time and leaving room for further optimization. This study presents a novel approach to reduce sampling time and conserve energy by extracting early prediction results from the intermediate layer of the network and integrating them with the final layer's predictions in a Bayesian fashion. Experimental evaluations conducted on image classification tasks using MNIST, CIFAR-10, and CIFAR-100 datasets demonstrate the efficacy of our proposed method when applied to VGGNets and ResNets models. Results indicate a substantial energy reduction of 38.8% in VGGNets and 48.0% in ResNets, illustrating the potential for achieving significant efficiency gains in spiking neural networks. These findings contribute to the ongoing research in enhancing the performance of SNNs, facilitating their deployment in resource-constrained environments. Our code is available on GitHub: https://github.com/hanebarla/BayesianSpikeFusion.

Visualization of a chorus structure in multiple frog species by a sound discrimination device.

We developed a sound discrimination device to identify and localize the species of nocturnal animals in their natural habitat. The sound discrimination device is equipped with a microphone, a light-emitting diode, and a band-pass filter. By tuning the center frequency of the filter to include a dominant frequency of the calls of a focal species, we enable the device to be illuminated only when detecting the calls of the focal species. In experiments in a laboratory room, we tuned the sound discrimination devices to detect the calls of Hyla japonica or Rhacophorus schlegelii and broadcast the frog calls from loudspeakers. By analyzing the illumination pattern of the devices, we successfully identified and localized the two kinds of sound sources. Next, we placed the sound discrimination devices in a field site where actual male frogs (H. japonica and R. schlegelii) produced sounds. The analysis of the illumination pattern demonstrates the efficacy of the developed devices in a natural environment and also enables us to extract pairs of male frogs that significantly overlapped or alternated their calls.

Visualizing Phonotactic Behavior of Female Frogs in Darkness.

Many animals use sounds produced by conspecifics for mate identification. Female insects and anuran amphibians, for instance, use acoustic cues to localize, orient toward and approach conspecific males prior to mating. Here we present a novel technique that utilizes multiple, distributed sound-indication devices and a miniature LED backpack to visualize and record the nocturnal phonotactic approach of females of the Australian orange-eyed tree frog (Litoria chloris) both in a laboratory arena and in the animal's natural habitat. Continuous high-definition digital recording of the LED coordinates provides automatic tracking of the female's position, and the illumination patterns of the sound-indication devices allow us to discriminate multiple sound sources including loudspeakers broadcasting calls as well as calls emitted by individual male frogs. This innovative methodology is widely applicable for the study of phonotaxis and spatial structures of acoustically communicating nocturnal animals.

Spatio-temporal dynamics in collective frog choruses examined by mathematical modeling and field observations.

This paper reports theoretical and experimental studies on spatio-temporal dynamics in the choruses of male Japanese tree frogs. First, we theoretically model their calling times and positions as a system of coupled mobile oscillators. Numerical simulation of the model as well as calculation of the order parameters show that the spatio-temporal dynamics exhibits bistability between two-cluster antisynchronization and wavy antisynchronization, by assuming that the frogs are attracted to the edge of a simple circular breeding site. Second, we change the shape of the breeding site from the circle to rectangles including a straight line, and evaluate the stability of two-cluster and wavy antisynchronization. Numerical simulation shows that two-cluster antisynchronization is more frequently observed than wavy antisynchronization. Finally, we recorded frog choruses at an actual paddy field using our sound-imaging method. Analysis of the video demonstrated a consistent result with the aforementioned simulation: namely, two-cluster antisynchronization was more frequently realized.