Small-scale neural simulation of Caenorhabditis elegans on Loihi-2 using NeuroML2Loihi

Danny Rosen; Ramashish Gaurav; Yang Yi

cs.NEOct 18, 2026

Small-scale neural simulation of Caenorhabditis elegans on Loihi-2 using NeuroML2Loihi

Authors: Danny Rosen, Ramashish Gaurav, Yang Yi

Organizations: Virginia Tech, Institute for Advanced Computing

Abstract

Neural simulation is an essential methodology for studying biological nervous systems, but it remains limited by the computational costs of repeatedly updating neural and synaptic states over many timesteps. Neuromorphic hardware is a promising substrate for this class of workloads because such devices combine event-driven execution, localized memory, and spike-based communication. However, neuromorphic software remains comparatively fragmented and difficult to integrate with the model-description standards used by the computational neuroscience community. This disconnect limits reproducible experimentation and makes direct comparison between conventional and neuromorphic simulation workflows difficult. This paper introduces ``NeuroML2Loihi'', an open-source framework for translating NeuroML model descriptions into Loihi~~2-deployable Lava primitives, enabling standardized biological models to be executed on Intel's Loihi~~2 neuromorphic chip. We evaluate our framework using c302-derived C. elegans neural models across three connectome-scale configurations. Cross-platform similarity is assessed using spike-train and voltage-trace metrics, while latency, power, and energy consumption are benchmarked against the conventional reference workflow. Across the evaluated models, Loihi2 execution enabled through NeuroML2Loihi achieves 514--1211 $\times$ lower latency and a 1.8--1.9 $\times$ reduction in mean wattage relative to the NEURON reference, producing substantially lower total energy consumption primarily through reduced execution time. These results demonstrate that standards-based neuromorphic neural simulation is feasible on Loihi2 and can support meaningful comparison against established computational neuroscience workflows.

Figures & tables

Panel of Figure 1 — Figure 1. The NeuroML2Loihi framework. A NeuroML model ( .nml ) and its LEMS simulation description ( .xml ) are read by separate parsers into a Network (a NetworkX graph) and a Configuration. Guided by the Mapping module, NeuroML components are assembled into Lava primitives (LIF/Dense) within the Network. The Experiment couples the Network and Configuration with an externally selected Backend—Loihi 2 hardware or the LoihiSim simulator—executes the run, and records outputs for Analysis.

Panel of Figure 2 — Figure 2. Vectorized Lava process network. All neurons are represented by a single LIF process connected recurrently to a Dense synaptic process through the adjacency matrix $W$ derived from the parsed NeuroML model.

Panel of Figure 3 — Figure 3. Hardware input and recording pipeline. Input spikes drive the vectorized LIF process, which carries a recurrent Dense synapse. Spike outputs are routed through a Dense process to Receiver and StateProbe processes, while membrane voltage is recorded directly by a StateProbe on the neuron’s $v$ property.

Panel of Figure 4 — Figure 4. Visualization of the Full ( A ) and Active ( B ) neuron subsets in the c302-derived NeuroML2 networks used in this study. Figures were generated using from the corresponding NeuroML model files used for simulation using pyNeuroML ( Vella et al., 2014 ) .

Parameter	Value
$d_{u}$	$49$
$d_{v}$	$8$
$v_{\mathrm{th}}$	$131,071$
Recurrent scale	$10\cdot 2^{5}=320$
Input scale	$4$
Biological duration	$300\text{\,}\mathrm{ms}$

Table 1. Model parameters.

metric	N2L vs NEURON	N2L vs Poisson	Ratio
VP@1	1.0914	3.8133	0.2862
VP@50	5.5700	12.2079	0.4563
VR	3.4194	4.7154	0.7252
$\lvert\Delta\mathrm{spikes}\rvert/\mathrm{neuron}$	1.0000	3.5000	0.2857

Table 2. NeuroML2Loihi vs NEURON

metric	PLM	Active	Full
Active Neurons Loihi 2	2	18	18
Active Neurons NEURON	2	13	13
voltage r	-0.1441	0.4334	0.4334
voltage RMSE ( $mV$ )	8.8322	3.9428	0.2350
voltage MAE ( $mV$ )	7.2744	3.3601	0.2003

Table 3. Voltage (NeuroML2Loihi affine-mapped to NEURON voltage scales)

Figure 5. Superimposed PLML voltage traces from the NeuroML2Loihi and NEURON simulations. The comparison illustrates that the platforms produce similar structured voltage evolution while differing in the exact subthreshold trajectory due to the use of Lava LIF and Dense primitives on Loihi versus the NeuroML/c302 neuron and synapse dynamics used by the reference simulator.

Model	N2L latency (s)	NEURON run time (s)	Speedup	N2L energy (J)	NEURON energy (J)	Energy (N2L / NEURON)
Minimal	$0.0243\pm 0.0001$	$12.47\pm 0.11$	$514\times$	$0.0122\pm 0.0001$	$11.65\pm 0.89$	0.105%
Active	$0.0242\pm 0.0001$	$18.62\pm 0.09$	$768\times$	$0.0123\pm 0.0001$	$18.11\pm 0.36$	0.0676%
Full	$0.0258\pm 0.0001$	$31.19\pm 0.12$	$1211\times$	$0.0131\pm 0.0001$	$28.52\pm 0.55$	0.0458%

Table 4. N2L on Loihi 2 vs NEURON on Jetson – Latency and Energy

Model	N2L power (W)	NEURON power (W)	Power ratio (NEURON / N2L)
Minimal	$0.5047\pm 0.0018$	$0.9393\pm 0.0706$	1.9x
Active	$0.5052\pm 0.0020$	$0.9778\pm 0.0199$	1.9x
Full	$0.5067\pm 0.0020$	$0.9189\pm 0.0178$	1.8x

Table 5. Power comparison between N2L simulation on Loihi-2 vs NEURON on Jetson

Model	Cores	Footprint (MB)	Total % mean
Minimal	3	0.001686	0.29966
Active	3	0.005424	0.96418
Full	9	0.6571	38.93860

Table 6. Loihi on-chip memory footprint

Neuromorphic Computing

Neuromorphic computing uses brain-inspired, event-driven algorithms and hardware to process information with low power and latency.

See all 186 papers

May 9, 2025

Sigma-Delta Neural Network Conversion on Loihi 2

Neuromorphic computing aims to improve the efficiency of artificial neural networks by taking inspiration from biological neurons and leveraging temporal sparsity, spatial sparsity, and compute near/in memory. Although these approaches have shown efficiency gains, training these spiking neural networks (SNN) remains difficult. The original attempts at converting trained conventional analog neural networks (ANN) to SNNs used the rate of binary spikes to represent neuron activations. This required many simulation time steps per inference, which degraded efficiency. Intel's Loihi 2 is a neuromorphic platform that supports graded spikes which can be used to represent changes in neuron activation. In this work, we use Loihi 2's graded spikes to develop a method for converting ANN networks to spiking networks, which exploits temporal and spatial sparsity. We evaluated the performance of this network on Loihi 2 and compared it to NVIDIA's Jetson Xavier edge AI platform. The results show that neuromorphic approaches achieve significant improvements in efficiency and latency (energy-delay product) over existing solutions.

Matthew Brehove, Sadia Anjum Tumpa, Espoir Kyubwa +2

ChromoLogic LLC Monrovia, California, USA · Pennsylvania State University University Park, Pennsylvania, USA

Efficient Neural Network Inference Spiking Neural Networks Neuromorphic Computing

Ask PDF

Aug 9, 2026

SuperNeuroMAT: An Efficient Matrix-based Simulator for Spiking Neural Networks

Spiking neural networks (SNNs) offer a promising pathway to energy-efficient AI and brain-inspired computing. However, their widespread adoption is hindered by a lack of fast, accessible, and versatile simulation frameworks. In this paper, we introduce SuperNeuroMAT, an open-source, scalable, and highly efficient Python-based SNN simulator. We devise a novel matrix-based approach to model the leaky integrate-and-fire (LIF) neuron dynamics and natively support dense and sparse execution modes. This enables fast simulation of approximately 10,000 neurons in dense mode and 100,000 neurons in sparse mode on standard laptops and desktops without requiring specialized hardware. We demonstrate that SuperNeuroMAT consistently outperforms four established SNN simulators---NEST, Brian2, BindsNET, and snnTorch---on two performance metrics (execution speed and peak resident memory) and across various network sizes and connection probabilities. Furthermore, we demonstrate SuperNeuroMAT's applicability across a diverse set of problems. SuperNeuroMAT can efficiently handle conventional machine learning benchmarks such as the Digits and citation network datasets as well as neuromorphic event-based vision tasks such as N-CARS and ASL-DVS. Moreover, it can be extended beyond machine learning workloads and facilitate general-purpose workloads. We validated this by implementing the neuromorphic shortest path algorithm and two arithmetic primitives (addition and multiplication). SuperNeuroMAT can be installed via the Python Package Index (PyPI), thereby lowering the barrier to entry into the field of neuromorphic computing and accelerating the broader development of neuromorphic algorithms.

Prasanna Date, Kevin Zhu, Shruti Kulkarni +12

Oak Ridge National Laboratory, Oak Ridge, 37830, Tennessee, United States. · George Mason University, Fairfax, 22030, Virginia, United States. · University of Tennessee, Knoxville, Knoxville, 37996, Tennessee, United States. +1

LIF Neurons Spiking Neural Networks Neuromorphic Computing

Ask PDF

Jun 7, 2026

Convolutional Sparse Coding via the Locally Competitive Algorithm on Loihi 2

Sparse coding provides a principled framework for signal representation by expressing an input as a linear combination of only a small number of basis functions. The Locally Competitive Algorithm (LCA) is particularly attractive in the context of neuromorphic computing because its dynamics, leaky integration, thresholding, and lateral inhibition map naturally to neuromorphic hardware. While prior work has studied non-convolutional LCA on Loihi 2, the convolutional setting is of particular interest because it introduces spatial structure, weight sharing, overlapping receptive fields, and scaling behavior that are more representative of practical sparse inference workloads. In this work, we present a Loihi 2 implementation of convolutional sparse coding via the LCA and evaluate it against a conventional GPU baseline on the same inference problems. The implementation follows a one-layer recurrent LCA formulation and extends it to convolutional feature maps with local inhibitory kernels derived from pairwise filter interactions. To the best of our knowledge, this is the first implementation and benchmark of convolutional LCA on Loihi 2. Our goal is not only to demonstrate feasibility, but also to clarify in which operating regimes convolutional sparse inference becomes attractive on neuromorphic hardware. The resulting study positions convolutional LCA as a useful benchmark for structured sparse inference on emerging neuromorphic systems.

Geoffrey Kasenbacher, Daniel Ruepp, Gerrit A. Ecke

Mercedes-Benz AG · Böblingen, Germany · Institut für Robotik und Kognitive Systeme, Universität zu Lübeck

Representation Learning Neuromorphic Computing Sparse Coding

Ask PDF

Abstract

Neuromorphic Computing

Explore similar work

Sigma-Delta Neural Network Conversion on Loihi 2

SuperNeuroMAT: An Efficient Matrix-based Simulator for Spiking Neural Networks

Convolutional Sparse Coding via the Locally Competitive Algorithm on Loihi 2