China
Africa
Explore chapters and articles related to this topic
Deep Learning
Published in Jan Žižka, František Dařena, Arnošt Svoboda, Text Mining with Machine Learning, 2019
Jan Žižka, František Dařena, Arnošt Svoboda
The R-language add-on system is no exception, and a variety of methods and algorithms can be found, sometimes more general, sometimes tailor-made for specific task groups. For ease of use and installation, this chapter confines itself to demonstrating the use of a single artificial neural network system – SNNS, which stands for Stuttgart Neural Network Simulator, one of the outstanding simulation systems [287].
Emerging memristive neurons for neuromorphic computing and sensing
Published in Science and Technology of Advanced Materials, 2023
Zhiyuan Li, Wei Tang, Beining Zhang, Rui Yang, Xiangshui Miao
In the era of AI, neuromorphic computing, a parallel computing architecture, has become a promising candidate for high-performance hardware systems. Scientists proposed artificial neurobiological network models (i.e. SNNs and ANNs) to develop neuromorphic computing, as illustrated in Figure 6. ANNs can perform supervised, semi-supervised, unsupervised, and reinforcement learning algorithms, and excel in deep learning tasks that had large amounts of computational resources for the training data [2,122]. They usually receive consecutive values and output consecutive values (i.e. floating point, fixed point or analog value), biologically inaccurate and do not mimic the abundant dynamics of biological spiking neurons. In contrast, SNNs, as third generation neural network, mimic the brain processes information more faithfully, in which the internal neurons communicate with each other through the sequence and timing binary spiking signals (i.e. a rate-coding or spatiotemporal-coding) [60,123,124]. When processing complex temporal intelligence tasks (e.g. event-driven information processing), SNNs can show great advantages over ANNs. However, traditional SNN-based hardware implementations usually were implemented by CMOS circuits and external waveform generators, which require more complex circuit design, higher power consumption, and larger chip space [16,52,125]. The emerging memristive devices provide new insights into highly efficient and compact SNNs hardware system. The nonlinearity and dynamics in the above artificial memristive neuron provide vital substrates for implementation of SNNs. In this section, we conduct a comprehensive review of building a spike-based neuromorphic hardware system exploiting memristive neuron.
Application of neural network in order to recognise individuality of course of vehicle and pedestrian body contacts during accidents
Published in International Journal of Crashworthiness, 2018
We assume that the ‘Injury coefficients’ bear the signature defining the type of a crash and are considered to be a part of its course similar to other circumstances arising therefrom. For the the identification (modelling), we used a forward neural network trained by the Backpropagation algorithm [4]. The Stuttgart Neural Network Simulator (SNNS, http://www-ra.informatik.uni-tuebingen.de/SNNS/) was used to train the neural network.