TripleBrain: An Edge Neuromorphic Architecture for High-accuracy Single-layer Spiking Neural Network with On-chip Self-organizing and Reinforcement Learning

This article presents the results of a numerical simulation and an experimental study of the electrical circuit of a hardware spiking perceptron based on a memristor-diode crossbar. That has required developing and manufacturing a measuring bench, the electrical circuit of which consists of a hardware perceptron circuit and an input peripheral electrical circuit to implement the activation functions of the neurons and ensure the operation of the memory matrix in a spiking mode. The authors have performed a study of the operation of the hardware spiking neural network with memristor synapses in the form of a memory matrix in the mode of a single-layer perceptron synapses. The perceptron can be considered as the first layer of a biomorphic neural network that performs primary processing of incoming information in a biomorphic neuroprocessor. The obtained experimental and simulation learning curves show the expected increase in the proportion of correct classifications with an increase in the number of training epochs. The authors demonstrate generating a new association during retraining caused by the presence of new input information. Comparison of the results of modeling and an experiment on training a small neural network with a small crossbar will allow creating adequate models of hardware neural networks with a large memristor-diode crossbar. The arrival of new unknown information at the input of the hardware spiking neural network can be related with the generation of new associations in the biomorphic neuroprocessor. With further improvement of the neural network, this information will be comprehended and, therefore, will allow the transition from weak to strong artificial intelligence.

Download Full-text

A Spiking Neural Network Based Autonomous Reinforcement Learning Model and Its Application in Decision Making

Advances in Brain Inspired Cognitive Systems - Lecture Notes in Computer Science ◽

10.1007/978-3-319-49685-6_12 ◽

2016 ◽

pp. 125-137

Author(s):

Guixiang Wang ◽

Yi Zeng ◽

Bo Xu

Keyword(s):

Neural Network ◽

Decision Making ◽

Reinforcement Learning ◽

Learning Model ◽

Spiking Neural Network ◽

Reinforcement Learning Model

Download Full-text

Control chart pattern clustering using a new self-organizing spiking neural network

Proceedings of the Institution of Mechanical Engineers Part B Journal of Engineering Manufacture ◽

10.1243/09544054jem1054 ◽

2008 ◽

Vol 222 (10) ◽

pp. 1201-1211 ◽

Cited By ~ 5

Author(s):

D T Pham ◽

M S Packianather ◽

E Y A Charles

Keyword(s):

Neural Network ◽

Control Chart ◽

Neural Network Model ◽

Spiking Neurons ◽

Spiking Neural Network ◽

Pattern Clustering ◽

Control Chart Patterns ◽

Chart Patterns ◽

Spiking Neural Network Model ◽

Self Organizing

This paper focuses on the architecture and learning algorithm associated with using a new self-organizing delay adaptation spiking neural network model for clustering control chart patterns. This temporal coding spiking neural network model employs a Hebbian-based rule to shift the connection delays instead of the previous approaches of delay selection. Here the tuned delays compensate the differences in the input firing times of temporal patterns and enables them to coincide. The coincidence detection capability of the spiking neuron has been utilized for pattern clustering. The structure of the network is similar to that of a Kohonen self-organizing map (SOM) except that the output layer neurons are coincidence detecting spiking neurons. An input pattern is represented by the neuron that is the first to fire among all the competing spiking neurons. Clusters within the input data are identified with the location of the winning neurons and their firing times. The proposed self-organized delay adaptation spiking neural network (SODA_SNN) has been utilized to cluster control chart patterns. The trained network obtained an average clustering accuracy of 96.1 per cent on previously unseen test data. This was achieved with a network of 8 × 8 spiking neurons trained for 20 epochs containing 1000 training examples. The improvement in clustering accuracy achieved by the proposed SODA_SNN on the unseen test data was twice as much as that on the training data when compared to the SOM.

Download Full-text