This work gives an introduction to SNNs and the underlying biological concepts. It gives an overview and comparison of some of the more commonly used SNN models. It discusses the scope of SNNs and some of the areas where they have been applied so far. Spiking neural networks or SNNs are inspired by the biological neuron. They are the next step towards the goal of replicating the mammalian brain in computational speed, efficiency and energy consumption.
First generation artificial neural networks (ANNs) or Perceptron use a [0,1] binary threshold function to approximate digital input and allow for linear classification. Second generation ANNs like multi-layer perceptron, feed-forward and recurrent neural networks use continuous activation functions like sigmoid which can approximate analog functions. Spiking neural networks, introduced by Hopfield in 1995, are third generation ANNs and aim at higher biological plausibilty than the first and second generations by including time intrinsically. They use the precise firing times of neurons to code information. SNNs are modelled on the biological neuron. It is therefore important to understand the basic biological concepts underlying SNNs.
Table of Contents
I. INTRODUCTION
A. Membrane Potential
B. Action Potential
C. Time as the basis of information coding
II. SPIKING NEURAL NETWORK MODELS
A. Threshold-Fire Models
1) Integrate-and-Fire (I&F) Model
2) Leaky-Integrate-and-Fire (LIF) Model
3) Spike Response Model (SRM)
B. Conductance-based Models
III. COMPARING SNN MODELS
IV. APPLICATIONS OF SNNS
A. Cognitive Hardware
B. Vision-based applications
C. Analysis of spatio-temporal data
D. Other areas
V. SUMMARY
Objectives and Topics
This work aims to provide a comprehensive introduction to Spiking Neural Networks (SNNs), exploring their biological foundations and comparing various common modeling approaches to evaluate their computational efficiency against biological plausibility.
- Biological principles of neurons and membrane potential
- Comparison of SNN models ranging from simple to complex
- The plausibility-efficiency tradeoff in neural modeling
- Real-world applications in hardware, vision, and data analysis
Excerpt from the book
A. Threshold-Fire Models
These models are based on the fundamental principle of biological neurons where a spike is generated when the membrane potential of the neuron crosses a certain threshold value from below.
1) Integrate-and-Fire (I&F) Model: This is the simplest SNN model and describes action potential as an event. This means that only the timing of spike is considered while the form of spike is ignored. Membrane potential is assumed as the integration of input spikes. These could either be multiple spikes of the same neuron; or spikes from multiple neurons in response to some stimulus; or both.
The following equation of the integrate-and-fire model is simply the derivative of the law of capacitance: C = Q/V . I(t) = C dV (t) / dt. When current I(t) causes summed potential V (t) to increase over time and cross threshold θ, a spike occurs. V (t) is then immediately reset to RMP and the process starts again. A summed potential lower than θ does not cause a spike and does not get reset. It is consequently retained till the next spike and does not decay. This is in contrast to the biological neuron, where a sub-threshold potential eventually decays to RMP. This lack of time-dependent memory is thus a limitation of this model, reducing its biological plausibility.
Summary of Chapters
I. INTRODUCTION: This chapter introduces Spiking Neural Networks (SNNs) as third-generation artificial neural networks and covers essential biological concepts such as membrane potential, action potentials, and temporal coding.
II. SPIKING NEURAL NETWORK MODELS: This section categorizes SNN models into Threshold-Fire models and Conductance-based models, detailing their mathematical formulations and biological relevance.
III. COMPARING SNN MODELS: This chapter discusses the trade-off between biological plausibility and computational efficiency in various SNN architectures.
IV. APPLICATIONS OF SNNS: This section outlines the practical utility of SNNs across fields like neuromorphic hardware, computer vision, and spatio-temporal data analysis.
V. SUMMARY: This final chapter synthesizes the core findings, highlighting the potential of SNNs for future technological advancements.
Keywords
Spiking Neural Networks, SNNs, Biological Neuron, Membrane Potential, Action Potential, Temporal Coding, Integrate-and-Fire Model, Leaky-Integrate-and-Fire, Spike Response Model, Hodgkin-Huxley Model, Neuromorphic Chips, Plausibility-Efficiency Tradeoff, Spatio-temporal Data, Pattern Recognition, Cognitive Hardware
Frequently Asked Questions
What is the primary focus of this work?
This work focuses on Spiking Neural Networks (SNNs) as a biologically inspired alternative to traditional neural networks, emphasizing their structural components and computational models.
What are the key thematic areas covered?
The key themes include the biological basis of neuronal signaling, the mathematical classification of SNN models, the trade-off between biological accuracy and processing speed, and practical application domains.
What is the central research question?
The work seeks to determine how SNNs can effectively mimic biological neurons while balancing the need for computational efficiency in hardware and software implementations.
Which scientific methods are employed?
The paper utilizes a comparative analysis approach, evaluating different mathematical models (e.g., I&F, LIF, SRM, HH) against their ability to represent biological processes and their resulting computational costs.
What does the main body address?
The main body systematically reviews the mechanics of membrane and action potentials, details various SNN architectures, and examines their performance in specific application sectors like vision and robotics.
Which keywords best describe the paper?
Relevant keywords include Spiking Neural Networks, temporal coding, neuromorphic hardware, biological plausibility, and computational efficiency.
How do SNNs differ from second-generation neural networks regarding time?
Unlike previous generations that use rate coding, SNNs incorporate time intrinsically by using the precise firing times of neurons to encode information, which is more biologically realistic.
What limitation does the basic Integrate-and-Fire model face?
The basic Integrate-and-Fire model lacks time-dependent memory because sub-threshold potentials do not decay to the resting membrane potential, which reduces its overall biological plausibility.
- Arbeit zitieren
- Garima Mittal (Autor:in), 2018, An Overview of Spiking Neural Networks, München, GRIN Verlag, https://www.grin.com/document/921629
