Evolving spiking neural networks for spatio-and spectro-temporal pattern recognition

Kasabov, Nikola

doi:10.1109/is.2012.6335110

Cited by 11 publications

(9 citation statements)

References 49 publications

(58 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Spiking neural networks (SNN) are advanced machine learning techniques. They simulate the behavior of biological neural networks by creating and updating connections between spiking neurons (synaptic connections) to learn temporal associations between them [1,2,[13][14][15][16][17][18][19][20][21]. SNN are considered the third generation of arti cial neural networks that have advantages over traditional neural networks due to the processing and incremental learning of temporal information.…”

Section: Spiking Neural Network (Snn)mentioning

confidence: 99%

Spiking Neural Networks for Predictive and Explainable Modelling of Multimodal Streaming Data with a Case Study on Financial Time-series and Online News

Kasabov

AbouHassan

Jagtap³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

Human intelligence is characterized by the ability to incrementally integrate different sources of information for a better decision making. This paper argues that brain-inspired spiking neural networks (SNN) can be used for predictive and explainable modelling of multimodal streaming data. The paper proposes a new method, based on the brain-inspired SNN architecture NeuCube, where, first, all streaming data are represented as numerical times series in the same time domain. Then a NeuCube model is incrementally trained on the integrated time series and continuously interpreted. The method is illustrated on integrated modelling of financial time series and online news. In contrast to traditional machine learning techniques, the proposed method reveals the dynamic interaction between all types of temporal variables and their impact on the model accuracy. The method is applicable on a wide range of multimodal time series, such as financial, medical, environmental, supporting also the use of massively parallel and low energy neuromorphic hardware.

show abstract

Section: Spiking Neural Network (Snn)mentioning

confidence: 99%

Spiking Neural Networks for Predictive and Explainable Modelling of Multimodal Streaming Data with a Case Study on Financial Time-series and Online News

Kasabov

AbouHassan

Jagtap³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…Although it is widely accepted that the graded activation of these analog neurons is equivalent to the firing rate of spiking neurons inside the biological neural network (Rueckauer et al 2017), the view that precise spike timing encodes information in the temporal domain that is important for neural computation is just as prominent in the neuroscience community (Gerstner and Kistler 2002;Gütig and Sompolinsky 2006). Consequently, the spiking neuron models are proposed to describe the dynamic of spike generation process, such that the additional temporal information including precise spike timing and phase could be better captured (Kasabov et al 2013).…”

Section: Introductionmentioning

confidence: 99%

“…One well-known single-spike learning algorithm is the Tempotron (Gütig and Sompolinsky 2006), whereby the output neuron is trained to fire a single spike in response to the correct input patterns and remain silent otherwise. The Rank-Order learning algorithms (Kasabov et al 2013) update synaptic weights based on the rank order of the arrival time of incoming spikes. Additionally, the time-to-first spike decoding scheme is employed at the output layer.…”

Section: Introductionmentioning

confidence: 99%

MPD-AL: An Efficient Membrane Potential Driven Aggregate-Label Learning Algorithm for Spiking Neurons

Zhang

Chua

et al. 2019

AAAI

View full text Add to dashboard Cite

One of the long-standing questions in biology and machine learning is how neural networks may learn important features from the input activities with a delayed feedback, commonly known as the temporal credit-assignment problem. The aggregate-label learning is proposed to resolve this problem by matching the spike count of a neuron with the magnitude of a feedback signal. However, the existing threshold-driven aggregate-label learning algorithms are computationally intensive, resulting in relatively low learning efficiency hence limiting their usability in practical applications. In order to address these limitations, we propose a novel membrane-potential driven aggregate-label learning algorithm, namely MPD-AL. With this algorithm, the easiest modifiable time instant is identified from membrane potential traces of the neuron, and guild the synaptic adaptation based on the presynaptic neurons’ contribution at this time instant. The experimental results demonstrate that the proposed algorithm enables the neurons to generate the desired number of spikes, and to detect useful clues embedded within unrelated spiking activities and background noise with a better learning efficiency over the state-of-the-art TDP1 and Multi-Spike Tempotron algorithms. Furthermore, we propose a data-driven dynamic decoding scheme for practical classification tasks, of which the aggregate labels are hard to define. This scheme effectively improves the classification accuracy of the aggregate-label learning algorithms as demonstrated on a speech recognition task.

show abstract

“…In contrast, spiking NNs can harness their inherent sensitivity to time to discover such temporal correlations by taking into account the relative timings between spikes. However, while various mechanisms, applications and implementations for spike-based processing have been studied in depth [5,7,12,13,15,21,22,24], surprisingly little attention has been paid to the problem of converting real-valued input into spike trains. Note that this is separate to the issue of encoding information using spike trains, which has been investigated in detail (e.g., see [29] for a survey of encoding techniques).…”

Section: Introductionmentioning

confidence: 99%

Adaptive conversion of real-valued input into spike trains

Hadjiivanov

2016

2016 International Joint Conference on Neural Networks (IJCNN)

View full text Add to dashboard Cite

This paper presents a biologically plausible method for converting real-valued input into spike trains for processing with spiking neural networks. The proposed method mimics the adaptive behaviour of retinal ganglion cells and allows input neurons to adapt their response to changes in the statistics of the input. Thus, rather than passively receiving values and forwarding them to the hidden and output layers, the input layer acts as a self-regulating filter which emphasises deviations from the average while allowing the input neurons to become effectively desensitised to the average itself. Another merit of the proposed method is that it requires only one input neuron per variable, rather than an entire population of neurons as in the case of the commonly used conversion method based on Gaussian receptive fields. In addition, since the statistics of the input emerge naturally over time, it becomes unnecessary to pre-process the data before feeding it to the network. This enables spiking neural networks to process raw, non-normalised streaming data. A proof-of-concept experiment is performed to demonstrate that the proposed method operates as expected.

show abstract

Evolving spiking neural networks for spatio-and spectro-temporal pattern recognition

Cited by 11 publications

References 49 publications

Spiking Neural Networks for Predictive and Explainable Modelling of Multimodal Streaming Data with a Case Study on Financial Time-series and Online News

Spiking Neural Networks for Predictive and Explainable Modelling of Multimodal Streaming Data with a Case Study on Financial Time-series and Online News

MPD-AL: An Efficient Membrane Potential Driven Aggregate-Label Learning Algorithm for Spiking Neurons

Adaptive conversion of real-valued input into spike trains

Contact Info

Product

Resources

About