Cascade Spiking Neuron Network For Event-based Image Classification In Noisy Environment

<div> <div> <div> <div> <p>Spiking Neuron Network (SNN) has shown advantages in processing event-based data for image classification. However, the classification accuracy of SNNs decreases in noisy environment. The cascade spiking neuron network (cascade-SNN) was proposed to solve this problem in this letter. We used spiking convolutional spiking neuron network (SCNN) for features extraction and liquid state machine (LSM) for read out. Compared with early works on ANNs, this network achieved the state-of-the-art classification accuracy in DVS-CIFAR10 dataset and DVS-Gesture dataset, which are both challenging dataset because of noisy environment. We conducted ablation experiments to verify the proposed structure is effective and analyzed the influence of different hyper-parameters. </p> </div> </div> </div> </div>

Cascade Spiking Neuron Network For Event-based Image Classification In Noisy Environment

<div> <div> <div> <div> <p>Spiking Neuron Network (SNN) has shown advantages in processing event-based data for image classification. However, the classification accuracy of SNNs decreases in noisy environment. The cascade spiking neuron network (cascade-SNN) was proposed to solve this problem in this letter. We used spiking convolutional spiking neuron network (SCNN) for features extraction and liquid state machine (LSM) for read out. Compared with early works on ANNs, this network achieved the state-of-the-art classification accuracy in DVS-CIFAR10 dataset and DVS-Gesture dataset, which are both challenging dataset because of noisy environment. We conducted ablation experiments to verify the proposed structure is effective and analyzed the influence of different hyper-parameters. </p> </div> </div> </div> </div>

Modeling Working Memory in a Spiking Neuron Network Accompanied by Astrocytes

We propose a novel biologically plausible computational model of working memory (WM) implemented by a spiking neuron network (SNN) interacting with a network of astrocytes. The SNN is modeled by synaptically coupled Izhikevich neurons with a non-specific architecture connection topology. Astrocytes generating calcium signals are connected by local gap junction diffusive couplings and interact with neurons via chemicals diffused in the extracellular space. Calcium elevations occur in response to the increased concentration of the neurotransmitter released by spiking neurons when a group of them fire coherently. In turn, gliotransmitters are released by activated astrocytes modulating the strength of the synaptic connections in the corresponding neuronal group. Input information is encoded as two-dimensional patterns of short applied current pulses stimulating neurons. The output is taken from frequencies of transient discharges of corresponding neurons. We show how a set of information patterns with quite significant overlapping areas can be uploaded into the neuron-astrocyte network and stored for several seconds. Information retrieval is organized by the application of a cue pattern representing one from the memory set distorted by noise. We found that successful retrieval with the level of the correlation between the recalled pattern and ideal pattern exceeding 90% is possible for the multi-item WM task. Having analyzed the dynamical mechanism of WM formation, we discovered that astrocytes operating at a time scale of a dozen of seconds can successfully store traces of neuronal activations corresponding to information patterns. In the retrieval stage, the astrocytic network selectively modulates synaptic connections in the SNN leading to successful recall. Information and dynamical characteristics of the proposed WM model agrees with classical concepts and other WM models.

Conversion-based Approach to Obtain an SNN Construction

Spiking Neuron Network (SNN) uses spike sequence for data processing, so it has an excellent characteristic of low power consumption. However, due to the immaturity of learning algorithm, the multiplayer network training has difficulty in convergence. Utilizing the mature learning algorithm and fast training speed of the back-propagation network, this paper proposes a method to converse the Convolutional Neural Network (CNN) to the SNN. First, the adjustment strategy for CNN is introduced. Then after training, the weight parameters in the model are extracted, which is the corresponding synaptic weight in the layer of the SNN. Finally, a new threshold-setting algorithm based on feedback is proposed to solve the critical problem of the threshold setting of neurons in the SNN. We evaluate our method on the CIFAR-10 datasets released by Hinton’s team. The experimental results show that the image classification accuracy of the SNN is more than 98% of that of CNN, and the theoretical value of power consumption per second is 3.9[Formula: see text]mW.