Search

The primary approaches used to train spiking neural networks (SNNs) involve either training artificial neural networks (ANNs) first and then transforming them into SNNs, or directly training SNNs using surrogate gradient techniques. Nevertheless, both of these methods encounter a shared challenge: they rely on frame-based methodologies, where asynchronous events are gathered into synchronous frames for computation. This strays from the authentic asynchronous, event-driven nature of SNNs, resulting in notable performance degradation when deploying the trained models on SNN simulators or hardware chips for real-time asynchronous computation. To eliminate this performance degradation, we propose a hardware-based SNN proxy learning method that is called Chip-In-Loop SNN Proxy Learning (CIL-SPL). This approach effectively eliminates the performance degradation caused by the mismatch between synchronous and asynchronous computations. To demonstrate the effectiveness of our method, we trained models using public datasets such as N-MNIST and tested them on the SNN simulator or hardware chip, comparing our results to those classical training methods.

Event-based dynamic vision sensors provide very sparse output in the form of spikes, which makes them suitable for low-power applications. Convolutional spiking neural networks model such event-based data and develop their full energy-saving potential when deployed on asynchronous neuromorphic hardware. Event-based vision being a nascent field, the sensitivity of spiking neural networks to potentially malicious adversarial attacks has received little attention so far. We show how white-box adversarial attack algorithms can be adapted to the discrete and sparse nature of event-based visual data, and demonstrate smaller perturbation magnitudes at higher success rates than the current state-of-the-art algorithms. For the first time, we also verify the effectiveness of these perturbations directly on neuromorphic hardware. Finally, we discuss the properties of the resulting perturbations, the effect of adversarial training as a defense strategy, and future directions.

Event-based dynamic vision sensors provide very sparse output in the form of spikes, which makes them suitable for low-power applications. Convolutional spiking neural networks model such event-based data and develop their full energy-saving potential when deployed on asynchronous neuromorphic hardware. Event-based vision being a nascent field, the sensitivity of spiking neural networks to potentially malicious adversarial attacks has received little attention so far. We show how white-box adversarial attack algorithms can be adapted to the discrete and sparse nature of event-based visual data, and demonstrate smaller perturbation magnitudes at higher success rates than the current state-of-the-art algorithms. For the first time, we also verify the eectiveness of these perturbations directly on neuromorphic hardware. Finally, we discuss the properties of the resulting perturbations, the eect of adversarial training as a defense strategy, and future direction. [ABSTRACT FROM AUTHOR]

Abstract Introduction Demyelination is a key factor in axonal degeneration and neural loss, leading to disability in multiple sclerosis (MS) patients. Transforming growth factor beta activated kinase 1 (TAK1) is a critical molecule involved in immune and inflammatory signaling pathways. Knockout of microglia TAK1 can inhibit autoimmune inflammation of the brain and spinal cord and improve the outcome of MS. However, it is unclear whether inhibiting TAK1 can alleviate demyelination. Methods Eight‐week‐old male c57bl/6j mice were randomly divided into five groups: (a) the control group, (b) the group treated with cuprizone (CPZ) only, (c) the group treated with 5Z‐7‐Oxozaenol (OZ) only, and (d) the group treated with both cuprizone and 15 μg/30 μg OZ. Demyelination in the mice of this study was induced by administration of CPZ (ig) at a daily dose of 400 mg/kg for consecutive 5 weeks. OZ was intraperitoneally administered at mentioned doses twice a week, starting from week 3 after beginning cuprizone treatment. Histology, rotarod test, grasping test, pole test, Western blot, RT‐PCR, and ELISA were used to evaluate corpus callosum demyelination, behavioral impairment, oligodendrocyte differentiation, TAK1 signaling pathway expression, microglia, and related cytokines. Results Our results demonstrated that OZ protected against myelin loss and behavior impairment caused by CPZ. Additionally, OZ rescued the loss of oligodendrocytes in CPZ‐induced mice. OZ inhibited the activation of JNK, p65, and p38 pathways, transformed M1 polarized microglia into M2 phenotype, and increased brain‐derived neurotrophic factor (BDNF) expression to attenuate demyelination in CPZ‐treated mice. Furthermore, OZ reduced the expression of proinflammatory cytokines and increases anti‐inflammatory cytokines in CPZ‐treated mice. Conclusion These findings suggest that inhibiting TAK1 may be an effective approach for treating demyelinating diseases.

Many studies have shown that solving addition and subtraction problems can induce overt shifts of spatial attention. In particular, right-side targets are detected faster than left-side targets when preceded by an addition operation, while left-side targets are detected faster than right-side targets when preceded by a subtraction operation. However, the interaction between space and arithmetic in multiplication or division is hardly studied and remains controversial. In order to make a strong case for the interaction between space and mental arithmetic, we attempted to replicate the spatial-arithmetic association in addition and subtraction (Experiment 1), and at the same time investigated whether shift of spatial attention would also be induced by multiplication or division operations (Experiment 2). We found that solving addition problems facilitated the detection of right-side targets, whereas left-side targets were detected faster after solving subtraction problems. However, no interaction between space and arithmetic operation was observed in multiplication or division. The implication of these findings is discussed.