Heterogeneous recurrent spiking neural network for spatio-temporal classification

Composing recurrent spiking neural networks using locally-recurrent motifs and risk-mitigating architectural optimization

In neural circuits, recurrent connectivity plays a crucial role in network function and stability. However, existing recurrent spiking neural networks (RSNNs) are often constructed by random connections without optimization. While RSNNs can produce rich dynamics that are critical for memory formation and learning, systemic architectural optimization of RSNNs is still an open challenge. We aim to enable systematic design of large RSNNs via a new scalable RSNN architecture and automated architectural optimization. We compose RSNNs based on a layer architecture called Sparsely-Connected Recurrent Motif Layer (SC-ML) that consists of multiple small recurrent motifs wired together by sparse lateral connections. The small size of the motifs and sparse inter-motif connectivity leads to an RSNN architecture scalable to large network sizes. We further propose a method called Hybrid Risk-Mitigating Architectural Search (HRMAS) to systematically optimize the topology of the proposed recurrent motifs and SC-ML layer architecture. HRMAS is an alternating two-step optimization process by which we mitigate the risk of network instability and performance degradation caused by architectural change by introducing a novel biologically-inspired "self-repairing" mechanism through intrinsic plasticity. The intrinsic plasticity is introduced to the second step of each HRMAS iteration and acts as unsupervised fast self-adaptation to structural and synaptic weight modifications introduced by the first step during the RSNN architectural "evolution." We demonstrate that the proposed automatic architecture optimization leads to significant performance gains over existing manually designed RSNNs: we achieve 96.44% on TI46-Alpha, 94.66% on N-TIDIGITS, 90.28% on DVS-Gesture, and 98.72% on N-MNIST. To the best of the authors' knowledge, this is the first work to perform systematic architecture optimization on RSNNs.

Co-learning synaptic delays, weights and adaptation in spiking neural networks

Spiking neural network (SNN) distinguish themselves from artificial neural network (ANN) because of their inherent temporal processing and spike-based computations, enabling a power-efficient implementation in neuromorphic hardware. In this study, we demonstrate that data processing with spiking neurons can be enhanced by co-learning the synaptic weights with two other biologically inspired neuronal features: (1) a set of parameters describing neuronal adaptation processes and (2) synaptic propagation delays. The former allows a spiking neuron to learn how to specifically react to incoming spikes based on its past. The trained adaptation parameters result in neuronal heterogeneity, which leads to a greater variety in available spike patterns and is also found in the brain. The latter enables to learn to explicitly correlate spike trains that are temporally distanced. Synaptic delays reflect the time an action potential requires to travel from one neuron to another. We show that each of the co-learned features separately leads to an improvement over the baseline SNN and that the combination of both leads to state-of-the-art SNN results on all speech recognition datasets investigated with a simple 2-hidden layer feed-forward network. Our SNN outperforms the benchmark ANN on the neuromorphic datasets (Spiking Heidelberg Digits and Spiking Speech Commands), even with fewer trainable parameters. On the 35-class Google Speech Commands dataset, our SNN also outperforms a GRU of similar size. Our study presents brain-inspired improvements in SNN that enable them to excel over an equivalent ANN of similar size on tasks with rich temporal dynamics.

Rapid diagnosis of systemic lupus erythematosus by Raman spectroscopy combined with spiking neural network

Multiple organs are affected by the autoimmune inflammatory connective tissue disease known as systemic lupus erythematosus (SLE). If not diagnosed and treated in a timely manner, it can lead to nephritis and damage to the blood system in severe cases, resulting in the patient's death. Therefore, correct and timely diagnosis and treatment are essential for patients. In this study, a framework based on neural network algorithm and Raman spectroscopy technique was established to diagnose SLE patients. Firstly, we pre-processed the obtained Raman data by three methods: baseline correction, smoothing processing and normalization methods, before using it as input for the model, and then ANN, ResNet and SNN classification models were established. The respective classification accuracies for SLE patients were 89.61%, 85.71%, and 95.65% for the three models, with corresponding AUC values of 0.8772, 0.8100, and 0.9555. The results of the experimental indicate that SNN possesses a good classification effect, and the number of model parameters is only 525,826, which is 414,221 less than that of ResNet model. Since the network only uses 0 and 1 to transmit information, and only has basic operations such as summation, compared with the second-generation artificial neural network, which simplifies the product operation of floating point numbers into multiple addition operations, the network has low energy consumption and is suitable for embedding portable Raman spectrometer for clinical diagnosis. This research highlights the significant potential for quick and precise SLE patient discrimination offered by Raman spectroscopy in conjunction with spiking neural networks.

Design and Development of an Imitation Detection System for Human Action Recognition Using Deep Learning

Human action recognition (HAR) is a rapidly growing field with numerous applications in various domains. HAR involves the development of algorithms and techniques to automatically identify and classify human actions from video data. Accurate recognition of human actions has significant implications in fields such as surveillance and sports analysis and in the health care domain. This paper presents a study on the design and development of an imitation detection system using an HAR algorithm based on deep learning. This study explores the use of deep learning models, such as a single-frame convolutional neural network (CNN) and pretrained VGG-16, for the accurate classification of human actions. The proposed models were evaluated using a benchmark dataset, KTH. The performance of these models was compared with that of classical classifiers, including K-Nearest Neighbors, Support Vector Machine, and Random Forest. The results showed that the VGG-16 model achieved higher accuracy than the single-frame CNN, with a 98% accuracy rate.