Spiking neural network explained

See main article: Artificial neural network. Spiking neural networks (SNNs) are artificial neural networks (ANN) that more closely mimic natural neural networks.[1] In addition to neuronal and synaptic state, SNNs incorporate the concept of time into their operating model. The idea is that neurons in the SNN do not transmit information at each propagation cycle (as it happens with typical multi-layer perceptron networks), but rather transmit information only when a membrane potential—an intrinsic quality of the neuron related to its membrane electrical charge—reaches a specific value, called the threshold. When the membrane potential reaches the threshold, the neuron fires, and generates a signal that travels to other neurons which, in turn, increase or decrease their potentials in response to this signal. A neuron model that fires at the moment of threshold crossing is also called a spiking neuron model.[2]

Although it was previously believed that the brain encoded information through spike rates, which can be considered as the analogue variable output of a traditional ANN,[3] research in the field of neurobiology has indicated that high speed processing cannot solely be performed through a rate based scheme. For example humans can perform an image recognition task at rate requiring no more than 10ms of processing time per neuron through the successive layers (going from the retina to the temporal lobe). This time window is too short for a rate based encoding. The precise spike timings in a small set of spiking neurons also has a higher information coding capacity compared with a rate based approach.[4]

The most prominent spiking neuron model is the leaky integrate-and-fire model.[5] In the integrate-and-fire model, the momentary activation level (modeled as a differential equation) is normally considered to be the neuron's state, with incoming spikes pushing this value higher or lower, until the state eventually either decays or—if the firing threshold is reached—the neuron fires. After firing, the state variable is reset to a lower value.

Various decoding methods exist for interpreting the outgoing spike train as a real-value number, relying on either the frequency of spikes (rate-code), the time-to-first-spike after stimulation, or the interval between spikes.

History

Many multi-layer artificial neural networks are fully connected, receiving input from every neuron in the previous layer and signalling every neuron in the subsequent layer. Although these networks have achieved breakthroughs in many fields, they are biologically inaccurate and do not mimic the operation mechanism of neurons in the brain of a living thing.

See also: Biological neuron model.

The biologically inspired Hodgkin–Huxley model of a spiking neuron was proposed in 1952. This model describes how action potentials are initiated and propagated. Communication between neurons, which requires the exchange of chemical neurotransmitters in the synaptic gap, is described in various models, such as the integrate-and-fire model, FitzHugh–Nagumo model (1961–1962), and Hindmarsh–Rose model (1984). The leaky integrate-and-fire model (or a derivative) is commonly used as it is easier to compute than the Hodgkin–Huxley model.[6]

While the notion of an artificial spiking neural network became very popular only during the first quarter of the twenty-first century, [7] [8] [9] there are a number of studies between 1980 and 1995 that supported the concept and in which the first models of this type of artificial neural networks appeared to simulate non-algorithmic intelligent information processing systems[10] [11] [12] . However, the very notion of the spiking neural network as a mathematical model had already been worked on in the early 1970s.[13]

Underpinnings

Information in the brain is represented as action potentials (neuron spikes), which may be grouped into spike trains or even coordinated waves of brain activity. A fundamental question of neuroscience is to determine whether neurons communicate by a rate or temporal code.[14] Temporal coding suggests that a single spiking neuron can replace hundreds of hidden units on a sigmoidal neural net.

An SNN computes in the continuous rather than the discrete domain. The idea is that neurons may not test for activation in every iteration of propagation (as is the case in a typical multilayer perceptron network), but only when their membrane potentials reach a certain value. When a neuron is activated, it produces a signal that is passed to connected neurons, raising or lowering their membrane potential.

In a spiking neural network, a neuron's current state is defined as its membrane potential (possibly modeled as a differential equation).[15] An input pulse causes the membrane potential to rise for a period of time and then gradually decline. Encoding schemes have been constructed to interpret these pulse sequences as a number, taking into account both pulse frequency and pulse interval. A neural network model based on pulse generation time can be established.[16] Using the exact time of pulse occurrence, a neural network can employ more information and offer better computing properties.[17]

The SNN approach produces a continuous output instead of the binary output of traditional artificial neural networks (ANNs). Pulse trains are not easily interpretable, hence the need for encoding schemes as above. However, a pulse train representation may be more suited for processing spatiotemporal data (or continual real-world sensory data classification).[18] SNNs consider space by connecting neurons only to nearby neurons so that they process input blocks separately (similar to CNN using filters). They consider time by encoding information as pulse trains so as not to lose information in a binary encoding. This avoids the additional complexity of a recurrent neural network (RNN). It turns out that impulse neurons are more powerful computational units than traditional artificial neurons.[19]

SNNs are theoretically more powerful than so called "second-generation networks" defined in as "[ANNs] based on computational units that apply activation function with a continuous set of possible output values to a weighted sum (or polynomial) of the inputs; however, SNN training issues and hardware requirements limit their use. Although unsupervised biologically inspired learning methods are available such as Hebbian learning and STDP, no effective supervised training method is suitable for SNNs that can provide better performance than second-generation networks. Spike-based activation of SNNs is not differentiable thus making it hard to develop gradient descent based training methods to perform error backpropagation.

SNNs have much larger computational costs for simulating realistic neural models than traditional ANNs.[20]

Pulse-coupled neural networks (PCNN) are often confused with SNNs. A PCNN can be seen as a kind of SNN.

Currently there are a few challenges when using SNNs that researchers are actively working on. The first challenge concerns the nondifferentiability of the spiking nonlinearity. The expressions for both the forward- and backward-learning methods contain the derivative of the neural activation function which is non-differentiable because neuron's output is either 1 when it spikes, and 0 otherwise. This all-or-nothing behavior of the binary spiking nonlinearity stops gradients from “flowing” and makes LIF neurons unsuitable for gradient-based optimization. The second challenge concerns the implementation of the optimization algorithm itself. Standard BP can be expensive in terms of computation, memory, and communication and may be poorly suited to the constraints dictated by the hardware that implements it (e.g., a computer, brain, or neuromorphic device).[21] Regarding the first challenge there are several approaches to resolving it. A few of them are:

  1. resorting to entirely biologically inspired local learning rules for the hidden units
  2. translating conventionally trained “rate-based” NNs to SNNs
  3. smoothing the network model to be continuously differentiable
  4. defining an SG (Surrogate Gradient) as a continuous relaxation of the real gradients

In the development of SNNs, incorporating additional neuron dynamics like Spike Frequency Adaptation (SFA) into neuron models marks a notable advance, enhancing both efficiency and computational power.[5] [22] These neurons stand in between biological complexity and computational complexity.[23] Originating from biological insights, SFA offers significant computational benefits by reducing power usage through efficient coding,[24] especially in cases of repetitive or intense stimuli. This adaptation improves signal clarity against background noise and introduces an elementary short-term memory at the neuron level, which in turn, refines the accuracy and efficiency of information processing.[25] Recently, this phenomenon was mostly achieved using compartmental neuron models. The simpler versions are of neuron models with adaptive thresholds, an indirect way of achieving SFA. It equips SNNs with improved learning capabilities, even with constrained synaptic plasticity, and elevates computational efficiency.[26] [27] This feature lessens the demand on network layers by decreasing the need for spike processing, thus cutting down on computational load and memory access time—essential aspects of neural computation. Moreover, SNNs utilizing neurons capable of SFA achieve levels of accuracy that rival those of conventional artificial neural networks, including those based on long short-term memory models,[28] while also requiring fewer neurons for comparable computational tasks. This efficiency not only streamlines the computational workflow but also conserves space and energy, offering a pragmatic step forward in the practical application of SNNs for complex computing tasks while maintaining a commitment to technical integrity.https://www.nature.com/articles/s41467-024-51110-5

Applications

SNNs can in principle be applied to the same applications as traditional ANNs.[29] In addition, SNNs can model the central nervous system of biological organisms, such as an insect seeking food without prior knowledge of the environment.[30] Due to their relative realism, they can be used to study the operation of biological neural circuits. Starting with a hypothesis about the topology of a biological neuronal circuit and its function, recordings of this circuit can be compared to the output of the corresponding SNN, evaluating the plausibility of the hypothesis. However, there is a lack of effective training mechanisms for SNNs, which can be inhibitory for some applications, including computer vision tasks.

As of 2019 SNNs lag behind ANNs in terms of accuracy, but the gap is decreasing, and has vanished on some tasks.[31]

When using SNNs for image based data, the images need to be converted into binary spike trains.[32] Types of encodings include:[33]

Software

A diverse range of application software can simulate SNNs. This software can be classified according to its uses:

SNN simulation

These simulate complex neural models with a high level of detail and accuracy. Large networks usually require lengthy processing. Candidates include:[34]

Hardware

Future neuromorphic architectures[37] will comprise billions of such nanosynapses, which require a clear understanding of the physical mechanisms responsible for plasticity. Experimental systems based on ferroelectric tunnel junctions have been used to show that STDP can be harnessed from heterogeneous polarization switching. Through combined scanning probe imaging, electrical transport and atomic-scale molecular dynamics, conductance variations can be modelled by nucleation-dominated reversal of domains. Simulations show that arrays of ferroelectric nanosynapses can autonomously learn to recognize patterns in a predictable way, opening the path towards unsupervised learning.[38]

Benchmarks

Classification capabilities of spiking networks trained according to unsupervised learning methods[43] have been tested on the common benchmark datasets, such as, Iris, Wisconsin Breast Cancer or Statlog Landsat dataset.[44] [45] Various approaches to information encoding and network design have been used. For example, a 2-layer feedforward network for data clustering and classification. Based on the idea proposed in Hopfield (1995) the authors implemented models of local receptive fields combining the properties of radial basis functions (RBF) and spiking neurons to convert input signals (classified data) having a floating-point representation into a spiking representation.[46] [47]

See also

Notes and References

  1. Maass W . Networks of spiking neurons: The third generation of neural network models. Neural Networks. 10. 9. 1997. 1659–1671. 0893-6080. 10.1016/S0893-6080(97)00011-7.
  2. Book: Gerstner W, Kistler WM . Spiking neuron models : single neurons, populations, plasticity. 2002. Cambridge University Press . 0-511-07817-X. Cambridge, U.K.. 57417395.
  3. Wang . Xiangwen . Lin . Xianghong . Dang . Xiaochao . 2020-05-01 . Supervised learning in spiking neural networks: A review of algorithms and evaluations . Neural Networks . 125 . 258–280 . 10.1016/j.neunet.2020.02.011 . 32146356 . 212638634 . 0893-6080.
  4. Taherkhani . Aboozar . Belatreche . Ammar . Li . Yuhua . Cosma . Georgina . Maguire . Liam P. . McGinnity . T. M. . 2020-02-01 . A review of learning in biologically plausible spiking neural networks . Neural Networks . 122 . 253–272 . 10.1016/j.neunet.2019.09.036 . 31726331 . 207904985 . 0893-6080.
  5. Ganguly . Chittotosh . Bezugam . Sai Sukruth . Abs . Elisabeth . Payvand . Melika . Dey . Sounak . Suri . Manan . 2024-02-01 . Spike frequency adaptation: bridging neural models and neuromorphic applications . Communications Engineering . en . 3 . 1 . 22 . 10.1038/s44172-024-00165-9 . 2731-3395. free . 11053160 .
  6. Book: Lee D, Lee G, Kwon D, Lee S, Kim Y, Kim J . 2018 ACM/IEEE 45th Annual International Symposium on Computer Architecture (ISCA) . Flexon: A Flexible Digital Neuron for Efficient Spiking Neural Network Simulations . June 2018 . 275–288 . 10.1109/isca.2018.00032. 978-1-5386-5984-7 . 50778421 .
  7. Goodman, D. F., & Brette, R. (2008). Brian: a simulator for spiking neural networks in python. Frontiers in neuroinformatics, 2, 350.
  8. Vreeken, J. (2003). Spiking neural networks, an introduction
  9. Yamazaki. K. . Vo-Ho. V. K.. Bulsara. D. . Le . N . 30 June 2022. Spiking neural networks and their applications: A review. Brain Sciences. 12. 7. 863. 10.3390/brainsci12070863. free . 35884670 . 9313413 .
  10. Ballard, D. H. (1987, July). Modular learning in neural networks. In Proceedings of the sixth National conference on Artificial intelligence-Volume 1 (pp. 279-284).
  11. Peretto, P. (1984). Collective properties of neural networks: a statistical physics approach. Biological cybernetics, 50(1), 51-62.
  12. Kurogi, S. (1987). A model of neural network for spatiotemporal pattern recognition. Biological cybernetics, 57(1), 103-114.
  13. Anderson, J. A. (1972). A simple neural network generating an interactive memory. Mathematical biosciences, 14(3-4), 197-220.
  14. Book: Wulfram Gerstner . Gerstner W . Spiking Neurons . Maass W, Bishop CM . 2001 . 978-0-262-63221-8 . Pulsed Neural Networks . MIT Press . https://books.google.com/books?id=jEug7sJXP2MC&q=%22Pulsed+Neural+Networks%22+rate-code+neuroscience&pg=PA3 .
  15. Hodgkin . A. L. . Huxley . A. F. . 1952-08-28 . A quantitative description of membrane current and its application to conduction and excitation in nerve . The Journal of Physiology . en . 117 . 4 . 500–544 . 10.1113/jphysiol.1952.sp004764 . 0022-3751 . 1392413 . 12991237.
  16. Dan . Yang . Poo . Mu-Ming . July 2006 . Spike Timing-Dependent Plasticity: From Synapse to Perception . Physiological Reviews . en . 86 . 3 . 1033–1048 . 10.1152/physrev.00030.2005 . 16816145 . 0031-9333.
  17. Nagornov . Nikolay N. . Lyakhov . Pavel A. . Bergerman . Maxim V. . Kalita . Diana I. . 2024 . Modern Trends in Improving the Technical Characteristics of Devices and Systems for Digital Image Processing . IEEE Access . 12 . 44659–44681 . 10.1109/ACCESS.2024.3381493 . 2024IEEEA..1244659N . 2169-3536. free .
  18. A robust modular spiking neural networks training methodology for time-series datasets: With a focus on gesture control. 2020. Van Wezel M . Master of Science . Delft University of Technology .
  19. Maass W . Networks of spiking neurons: The third generation of neural network models . Neural Networks . 10 . 9 . 1997 . 1659–1671 . 10.1016/S0893-6080(97)00011-7.
  20. Furber . Steve . August 2016 . Large-scale neuromorphic computing systems . Journal of Neural Engineering . en . 13 . 5 . 051001 . 10.1088/1741-2560/13/5/051001 . 27529195 . 2016JNEng..13e1001F . 1741-2552.
  21. Surrogate Gradient Learning in Spiking Neural Networks: Bringing the Power of Gradient-Based Optimization to Spiking Neural Networks . IEEE Signal Processing Magazine . 10.1109/msp.2019.2931595 . 2019 . Neftci . Emre O. . Mostafa . Hesham . Zenke . Friedemann . 36 . 6 . 51–63 . 2019ISPM...36f..51N . free .
  22. Salaj . Darjan . Subramoney . Anand . Kraisnikovic . Ceca . Bellec . Guillaume . Legenstein . Robert . Maass . Wolfgang . 2021-07-26 . O'Leary . Timothy . Behrens . Timothy E . Gutierrez . Gabrielle . Spike frequency adaptation supports network computations on temporally dispersed information . eLife . 10 . e65459 . 10.7554/eLife.65459 . free . 2050-084X . 8313230 . 34310281.
  23. Which model to use for cortical spiking neurons? . 2024-02-14 . IEEE Transactions on Neural Networks . 10.1109/tnn.2004.832719 . 2004 . Izhikevich . E.M. . 15 . 5 . 1063–1070 . 15484883 . 7354646 .
  24. Adibi, M., McDonald, J. S., Clifford, C. W. & Arabzadeh, E. Adaptation improves neural coding efficiency despite increasing correlations in variability. J. Neurosci. 33, 2108–2120 (2013)
  25. Laughlin . S. . 1981 . A simple coding procedure enhances a neuron's information capacity . Zeitschrift für Naturforschung C . 36 . 9–10 . 910–912 . 0341-0382 . 7303823.
  26. Immunity to Device Variations in a Spiking Neural Network With Memristive Nanodevices . 2024-02-14 . IEEE Transactions on Nanotechnology. 2013 . 10.1109/TNANO.2013.2250995 . 14416573 . Querlioz . Damien . Bichler . Olivier . Dollfus . Philippe . Gamrat . Christian . 12 . 3 . 288–295 . 2013ITNan..12..288Q .
  27. Yamazaki . Kashu . Vo-Ho . Viet-Khoa . Bulsara . Darshan . Le . Ngan . July 2022 . Spiking Neural Networks and Their Applications: A Review . Brain Sciences . en . 12 . 7 . 863 . 10.3390/brainsci12070863 . free . 2076-3425 . 9313413 . 35884670.
  28. Shaban . Ahmed . Bezugam . Sai Sukruth . Suri . Manan . 2021-07-09 . An adaptive threshold neuron for recurrent spiking neural networks with nanodevice hardware implementation . Nature Communications . en . 12 . 1 . 4234 . 10.1038/s41467-021-24427-8 . 34244491 . 8270926 . 2021NatCo..12.4234S . 2041-1723.
  29. Alnajjar F, Murase K . A simple Aplysia-like spiking neural network to generate adaptive behavior in autonomous robots . Adaptive Behavior . 14 . 5 . 306–324 . 10.1177/1059712308093869. 2008 . 16577867 .
  30. Book: Zhang X, Xu Z, Henriquez C, Ferrari S . 52nd IEEE Conference on Decision and Control . Spike-based indirect training of a spiking neural network-controlled virtual insect . 6798–6805 . Dec 2013 . 10.1109/CDC.2013.6760966. 978-1-4673-5717-3 . 10.1.1.671.6351 . 13992150 .
  31. Tavanaei A, Ghodrati M, Kheradpisheh SR, Masquelier T, Maida A . Deep learning in spiking neural networks . Neural Networks . 111 . 47–63 . March 2019 . 30682710 . 10.1016/j.neunet.2018.12.002 . 1804.08150 . 5039751 .
  32. Yamazaki K, Vo-Ho VK, Bulsara D, Le N . Spiking Neural Networks and Their Applications: A Review . Brain Sciences . 12 . 7 . 863 . June 2022 . 35884670 . 9313413 . 10.3390/brainsci12070863 . free .
  33. Kim Y, Park H, Moitra A, Bhattacharjee A, Venkatesha Y, Panda P . 2022-01-31 . Rate Coding or Direct Coding: Which One is Better for Accurate, Robust, and Energy-efficient Spiking Neural Networks? . cs.NE . 2202.03133.
  34. Abbott LF, Nelson SB . Synaptic plasticity: taming the beast . Nature Neuroscience . 3 . S11 . 1178–1183 . November 2000 . 11127835 . 10.1038/81453 . 2048100 .
  35. Atiya AF, Parlos AG . New results on recurrent network training: unifying the algorithms and accelerating convergence . IEEE Transactions on Neural Networks . 11 . 3 . 697–709 . May 2000 . 18249797 . 10.1109/72.846741 .
  36. Sanaullah S, Koravuna S, Rückert U, Jungeblut T . Evaluation of Spiking Neural Nets-Based Image Classification Using the Runtime Simulator RAVSim . International Journal of Neural Systems . 33 . 9 . 2350044 . August 2023 . 37604777 . 10.1142/S0129065723500442 . 259445644 .
  37. Sutton RS, Barto AG (2002) Reinforcement Learning: An Introduction. Bradford Books, MIT Press, Cambridge, MA.
  38. Boyn S, Grollier J, Lecerf G, Xu B, Locatelli N, Fusil S, Girod S, Carrétéro C, Garcia K, Xavier S, Tomas J, Bellaiche L, Bibes M, Barthélémy A, Saïghi S, Garcia V . 6 . Learning through ferroelectric domain dynamics in solid-state synapses . Nature Communications . 8 . 14736 . April 2017 . 28368007 . 5382254 . 10.1038/ncomms14736 . 2017NatCo...814736B .
  39. Book: 10.1109/IJCNN.2008.4634194 . Jin X, Furber SB, Woods JV . Efficient modelling of spiking neural networks on a scalable chip multiprocessor . 2008 IEEE International Joint Conference on Neural Networks (IEEE World Congress on Computational Intelligence) . 2812–2819 . 2008 . 978-1-4244-1820-6 . 2103654 .
  40. Web site: Neuromorphic Computing . Human Brain Project .
  41. Web site: Hardware: Available Systems. Human Brain Project. 2020-05-10.
  42. Web site: Markoff J . A new chip functions like a brain, IBM says . The New York Times . 8 August 2014 . B1 .
  43. Ponulak F, Kasiński A . Supervised learning in spiking neural networks with ReSuMe: sequence learning, classification, and spike shifting . Neural Computation . 22 . 2 . 467–510 . February 2010 . 19842989 . 10.1162/neco.2009.11-08-901 . 12572538 .
  44. Web site: Newman D, Hettich S, Blake C, Merz C . 1998 . UCI repository of machine learning databases. .
  45. Bohte S, Kok JN, La Poutré H . 2002 . Error-backpropagation in temporally encoded networks of spiking neurons. . Neurocomputing . 48 . 1–4 . 17–37 . 10.1016/S0925-2312(01)00658-0 .
  46. Pfister JP, Toyoizumi T, Barber D, Gerstner W . Optimal spike-timing-dependent plasticity for precise action potential firing in supervised learning . Neural Computation . 18 . 6 . 1318–1348 . June 2006 . 16764506 . 10.1162/neco.2006.18.6.1318 . 2005q.bio.....2037P . q-bio/0502037 . 6379045 .
  47. Bohte SM, La Poutré H, Kok JN . Unsupervised clustering with spiking neurons by sparse temporal coding and multilayer RBF networks. . IEEE Transactions on Neural Networks . March 2002 . 13 . 2. 426–435 . 10.1109/72.991428 . 18244443 .