Abstract:As numerical simulations grow in size and complexity, they become increasingly resource-intensive in terms of time and energy. While specialized hardware accelerators often provide order-of-magnitude gains and are state of the art in other scientific fields, their availability and applicability in computational neuroscience is still limited. In this field, neuromorphic accelerators, particularly mixed-signal architectures like the BrainScaleS systems, offer the most significant performance benefits. These systems maintain a constant, accelerated emulation speed independent of network model and size. This is especially beneficial when traditional simulators reach their limits, such as when modeling complex neuron dynamics, incorporating plasticity mechanisms, or running long or repetitive experiments. However, the analog nature of these systems introduces new challenges. In this paper we demonstrate the capabilities and advantages of the BrainScaleS-1 system and how it can be used in combination with conventional software simulations. We report the emulation time and energy consumption for two biologically inspired networks adapted to the neuromorphic hardware substrate: a balanced random network based on Brunel and the cortical microcircuit from Potjans and Diesmann.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: as the scale and complexity of numerical simulations keep increasing, their consumption of time and energy resources also becomes more and more intensive. Although dedicated hardware accelerators have provided performance improvements by orders of magnitude in other scientific fields, in the field of computational neuroscience, the availability and applicability of these accelerators are still limited. Specifically, this paper aims to demonstrate the significant performance advantages of neuromorphic hardware, especially mixed - signal architectures such as the BrainScaleS system, when simulating large - scale biologically - inspired spiking neural networks. To illustrate this problem specifically, the paper uses two biologically - inspired network models - the Balanced Random Network and the Cortical Microcircuit Model - to demonstrate the functions and advantages of the BrainScaleS - 1 system. These two models are adjusted to adapt to the constraints of neuromorphic hardware and are simulated on this hardware. The researchers evaluate the time efficiency and energy consumption of the simulation and discuss the advantages of analog neuromorphic systems and the practical application value of their combined use with traditional numerical simulators. ### Specific problem description 1. **Resource - intensive simulation**: As the simulation scale and complexity increase, the consumption of computational resources (time, energy) by traditional numerical simulation methods becomes very large. 2. **Application limitations of hardware accelerators**: Although dedicated hardware accelerators perform excellently in other fields, their application in computational neuroscience is still limited. 3. **Potential of neuromorphic hardware**: Explore whether neuromorphic chips, especially mixed - signal architectures (such as the BrainScaleS system), can provide significant performance improvements and solve the resource bottlenecks faced by traditional simulation methods. ### Core objectives of the paper - **Demonstrate performance advantages**: Prove the speed and energy - efficiency advantages of the BrainScaleS - 1 system in simulating complex neural networks through specific experimental results. - **Solve practical problems**: Explore how to use neuromorphic hardware to overcome the limitations of traditional simulation methods, especially in dealing with complex neural dynamics, plasticity mechanisms, and performance in long - term or repeated experiments. - **Promote cross - platform collaboration**: Emphasize the importance of the combined use of neuromorphic hardware and traditional numerical simulators to give full play to their respective advantages and promote the development of computational neuroscience. Through these efforts, the paper hopes to provide a more efficient and more potential computational tool for future neuroscience research.

Demonstrating the Advantages of Analog Wafer-Scale Neuromorphic Hardware

The BrainScaleS-2 Accelerated Neuromorphic System With Hybrid Plasticity

From Clean Room to Machine Room: Commissioning of the First-Generation BrainScaleS Wafer-Scale Neuromorphic System

Scalable Network Emulation on Analog Neuromorphic Hardware

An Accelerated Analog Neuromorphic Hardware System Emulating NMDA- and Calcium-Based Non-Linear Dendrites

Verification and Design Methods for the BrainScaleS Neuromorphic Hardware System

Real-Time Cortical Simulation on Neuromorphic Hardware

A Realistic Simulation Framework for Analog/Digital Neuromorphic Architectures

Multi-timescale synaptic plasticity on analog neuromorphic hardware

Extensible neuromorphic computing simulator based on a programmable hardware

Breaking Liebig's Law: An Advanced Multipurpose Neuromorphic Engine

BrainS: Customized multi-core embedded multiple scale neuromorphic system

Large-Scale Neuromorphic Spiking Array Processors: A Quest to Mimic the Brain

Advancing Neuromorphic Computing: Mixed-Signal Design Techniques Leveraging Brain Code Units and Fundamental Code Units

A System-on-Chip Based Hybrid Neuromorphic Compute Node Architecture for Reproducible Hyper-Real-Time Simulations of Spiking Neural Networks

How can neuromorphic hardware attain brain-like functional capabilities?

DYNAP-SE2: a scalable multi-core dynamic neuromorphic asynchronous spiking neural network processor

SuperNeuro: A Fast and Scalable Simulator for Neuromorphic Computing

Neuromorphic electronic circuits for building autonomous cognitive systems

To Spike or Not To Spike: A Digital Hardware Perspective on Deep Learning Acceleration

Scalable Digital Neuromorphic Architecture for Large-Scale Biophysically Meaningful Neural Network With Multi-Compartment Neurons