Explore chapters and articles related to this topic
Shallow Neural Networks
Published in Adrian A. Hopgood, Intelligent Systems for Engineers and Scientists, 2021
Spiking neural networks (SNNs) have been presented as a third generation of artificial neural networks that are intended as more biologically plausible models of neural processing. A key initiative has been the development of spiking neurons that communicate via the timing of spikes or a sequence of spikes (Ghosh-Dastidar and Adeli 2009). As with the first-generation neurons, spiking neurons produce only a binary output (0 or 1). The inputs and outputs are short-lived spikes whose timing conveys information, as in biological neurons, while the magnitude of the input and output spikes is immaterial.
The Memristor and its Implementation in Deep Neural Network Designing
Published in Rashmi Priyadarshini, R M Mehra, Amit Sehgal, Prabhu Jyot Singh, Artificial Intelligence, 2023
Melaku Nigus Getachew, Rashmi Priyadarshini, R. M. Mehra
SNNs are implemented in hardware in different forms. These implementations use different kinds of neuron models depending on the types of application for which the neural network is intended to be used. For example, if one has plan to form SNN network for direct investigations of neuroscience problem in silico test bed, then rather than bi-dimentional generalized adaptive integrate and fire nuron models, a more biologically- plausiable membrane ion-channel conductance-based neuron circuit models have to be used to design a hardware SNN network or neuromorphic system. Similar to biologicalneural networks, in spike neural networks, spikes are implemented for informationprocessing. In SNNs’ dynamics modeling it is not only the neurons’ state and weight that are considered but also the spike timing between the pre-and postsynaptic neuron firing. Out of several artificial neural networks, or deep neural networks forthat matter, the spiking neural network is the one which is built by mimicking thenetwork of the biological brain where nerve cells are transmitting information usingspikes and making connection with each other via synapse. For the hardware realization of a working SNN, memristor deviance is a promising candidate for itsimplementation as a synapse in the network. In SNN the weight of a synapse inthe network is updated using the STDP learning rule; by using this learning rule both unsupervised and supervised, machine learning methods can be demonstrated more efficiently [300]. SNN is now being used for numerous applications in different field. The SNN circuit is very complex and itspractical implementation requires further simplification of the network structureby taking various factors into considerations, such as the network area occupation, and power efficiency. A memristive deep neural network can be described from the perspective of the algorithm type the hardware neural network used during the traning cycle, from the perspective of the network architecture used during the design stage, and from that of the network circuit level implementation.
Oxide Memristor and Applications
Published in Simon Deleonibus, Emerging Devices for Low-Power and High-Performance Nanosystems, 2018
Mingyi Rao, Rivu Midya, J. Joshua Yang
A spiking neural network (SNN) is generally considered to be any artificial implementation of synaptic functions using asynchronous spikes of identical amplitude and duration [92]. Hence, a SNN is a much closer realization of a biological neural network than a threshold logic unit (TLU)-based ANN.
Hybrid feature learning framework for the classification of encrypted network traffic
Published in Connection Science, 2023
With the rising demand for Network traffic classification, a lot of studies have incorporated their models with convolutional spiking neural networks. A recent study indicated in Kumar and Sharma (2016), Lee et al. (2015), Liu et al. (2019) and Park et al. (2008) involve the use of the same model. The spiking neural networks have shown promising results in wide areas of applications which include detection, computation and recognition tasks. These models have been introduced in signal processing problems and have a wider scope in understanding the dynamic behaviour of the data packets. A comparitive study is given in table 2 and followed by the challenges in recent studies are tabulated in table 3.