Frontiers in Neuroscience · 2019 · 22 citations · 50 references
Life SciencesBiologyNeuroscienceComputational NeuroscienceNeural EngineeringSynaptic PlasticityNeural Circuit ModelingNeural ModelingNeural NetworksNeural System ModelingSpiking Neural NetworksStructural Synaptic PlasticityDendritic SpinesIonic Liquid MediaNeuron DynamicsSpiking ActivitySpiking Neural Network ModelsSpiking Neuron ModelsTemporal CodingTopological Features
One of the biggest struggles while working with artificial neural networks is being able to come up with models which closely match biological observations. In this paper, a new model for reservoir networks consisting of spiking neurons is introduced. Biological observations show that applying a stimulus to a network of neurons results in the creation of new dendritic spines which leads to new synapses being created. The latter forms the basis of the present study in which a new model for a spiking neural network in ionic liquid space is introduced. High plasticity of this model makes learning possible with a lesser number of neurons. In order to study the effect of the applied stimulus in ionic liquid space through time, a diffusion operator is used which somehow compensates for the separation between spatial and temporal coding in spiking neural networks and therefore, makes the mentioned model suitable for spatiotemporal patterns. Inspired by partial structural changes of human brain through the years, the proposed model evolves during learning process. The effect of evolution on the proposed model's performance is studied in this paper. Hence, in this paper, the topology of the proposed model is optimized for a classification problem using an evolutionary algorithm as well as some experiments. The N-MNIST dataset has been used to evaluate the performance of the proposed model. Similar to all reservoir networks, in order to classify this dataset, a readout layer is added to the model. Classification of the N-MNIST dataset using this model and without optimization results in a maximum accuracy of 7.69%, while a maximum accuracy of 98.38% is obtained when the evolutionary algorithm is applied.
50
Gradient-based learning applied to document recognition
Yann LeCun, Léon Bottou, Yoshua Bengio et al. · Proceedings of the IEEE · 1998 · 58.2K citations · Full text
Simple model of spiking neurons
Eugene M. Izhikevich · IEEE Transactions on Neural Networks · 2003 · 4.8K citations
A scaled conjugate gradient algorithm for fast supervised learning
Martin F. Møller · Neural Networks · 1993 · 4.1K citations