China 
Africa 
Explore chapters and articles related to this topic
Neuromorphic Nanoelectronics
Published in Klaus D. Sattler, st Century Nanoscience – A Handbook, 2020
Memristors challenge the very nature of computing as it has developed, based on the architecture of transistors acting as binary switches. However, the computing architecture of the brain provides another possible model—rather than digital logic, the brain runs on analog connections between neurons and synapses, which are made more or less conductive by stimulation, leading to the repeated firing of the synapse. This is the physical process underpinning learning at the neurological level, and is generally described as a change in “synaptic weight,” which over time can make synapses biochemically easier (or more difficult) to fire.14 This efficient approach to computation leads to the brain’s relatively low power consumption (20 W) and fault tolerance deriving from plasticity.15
H
Published in Phillip A. Laplante, Dictionary of Computer Science, Engineering, and Technology, 2017
Hebbian learning a method of modifying synaptic weights such that the strength of a synaptic weight is increased if both the presynaptic neuron and postsynaptic neuron are active simultaneously (synchronously) and decreased if the activity is asynchronous. In artificial neural networks, the synaptic weight is modified by a function of the correlation between the two activities.
R-STDP Based Spiking Neural Network for Human Action Recognition
Published in Applied Artificial Intelligence, 2020
In contrast to traditional neural networks, spiking neural network (SNN) (Meng, Jin, and Yin 2011) is a powerful tool in exploiting the temporal information. In addition, hundreds of neurons are integrated into a single on-chip node, in order to improve parallel processing, communication cost, and energy saving (Xiang and Meng 2018). The SNN generates binary neural output pulses called spikes and the information regarding the action sequence is expressed in terms of the relative timing of the spikes rather than the shape of the spikes. In this work, the R-STDP-based SNN used for the task of object recognition task is re-purposed to the task of action recognition. Initially, the directional features are extracted using gradient filters of different orientations. Then, the synaptic weight is modulated using the learning algorithm called (R-STDP) (Mozafari et al. 2018). The R-STDP uses reinforcement learning where the polarity of the synaptic plasticity is updated based on the reward/punishment signal. Finally, the network learns the shape of the specific pattern of the action sequence when the same pattern is subjected to the network several times.
Building memory devices from biocomposite electronic materials
Published in Science and Technology of Advanced Materials, 2020
Xuechao Xing, Meng Chen, Yue Gong, Ziyu Lv, Su-Ting Han, Ye Zhou
Biologically, neural memory is regarded to be closely related to synaptic weight and the strength of synaptic connections. In 2019, Liu and co-workers demonstrated new mesoscopic bioelectronic hybrid materials of silk fibroin (SF)-Ag nanoclusters (AgNCs@BSA; BSA: bovine serum albumin) based two terminal memristor-based synaptic devices [11]. Using AgNCs@BSA as an electron potential well enhanced the transport behavior of electricity in the SF film, which made the switching speed of the SF memory resistor significantly increased, that also showed unique synaptic performance and synaptic learning ability. In the connection between neuron synapses, the concentration of ionic species limited the activation/inhibition of the release of neurotransmitters and receptors for a certain period of time, thereby controlling the increase and decrease of synaptic weights (Figure 7(b)). The authors made use of the electrical conductivity of the two devices as a synaptic weight, and obtained a behavior similar to the nonlinear transmission characteristics of biological synapses. In addition, the author applied 250 ns width and 1.5 V positive and negative pulses to the device, and obtained the response of the change in synaptic weight. As the positive pulses increased, the conductance was enhanced, conversely, by changing the polarity of the pulses to the opposite, the conductance of the device decreases as the number of pulses increase (Figure 7(c)). In biological synapses, the synaptic weight increases as the interval between stimuli decreases, and this change is due to the increase in presynaptic Ca2+ concentration leading to the release of synaptic neurotransmission, what is more, this behavior is similar to that of PPF in biological synapses [106]. PPF is defined as
Mechanical property prediction of SPS processed GNP/PLA polymer nanocomposite using artificial neural network
Published in Cogent Engineering, 2020
O. T. Adesina, T. Jamiru, I. A. Daniyan, E. R. Sadiku, O. F. Ogunbiyi, O. S. Adesina, L. W. Beneke
The operations of ANN have its origin from the workings of the human nervous systems. It acquires knowledge like the human brain in a learning process and stores this knowledge via the interneuron connection known as synaptic weight. The ANN structure majorly divided into three segments: input layer, hidden layer, and output layer (Mahdavi Jafari, Soroushian, & Khayati, 2017). The structure in an ANN can be represented as Equation (1).