SpikeFit: Towards Optimal Deployment of Spiking Networks on Neuromorphic Hardware
Not on the map. Why?
Abstract (arXiv)
This paper introduces SpikeFit, a novel training method for Spiking Neural Networks (SNNs) that enables efficient inference on neuromorphic hardware, considering all its stringent requirements: the number of neurons and synapses that can fit on a single device, and lower bit-width representations (e.g., 4-bit, 8-bit). Unlike conventional compressing approaches that address only a subset of these requirements (limited numerical precision and limited number of neurons in the network), SpikeFit treats the allowed weights' discrete values themselves as learnable parameters co-optimized with the model, allowing for optimal Clusterization-Aware Training (CAT) of the model's weights at low precision (2-, 4-, or 8-bit) which results in higher network compression efficiency, as well as limiting the number of unique synaptic connections to a value required by neuromorphic processor. This joint optimization allows SpikeFit to find a discrete weight set aligned with hardware constraints, enabling the most complete deployment across a broader range of neuromorphic processors than existing methods of SNN compression support. Moreover, SpikeFit introduces a new hardware-friendly Fisher Spike Contribution (FSC) pruning method showing the state-of-the-art performance. We demonstrate that for spiking neural networks constrained to only four unique synaptic weight values (M = 4), our SpikeFit method not only outperforms state-of-the-art SNNs compression methods and conventional baselines combining extreme quantization schemes and clustering algorithms, but also meets a wider range of neuromorphic hardware requirements and provides the lowest energy use in experiments.
Key facts. Cited by 2 papers in the map (data through 7 Oct 2026). Builds on 24 of its 47 references that are in the map. First posted on arXiv 17 Oct 2025.
Builds on
24 of 47 references.
References · 24
- 2017Neural Discrete Representation Learningcited by 4,650
- 2013Estimating or Propagating Gradients Through Stochastic Neurons for Conditional Computationcited by 2,031
- 2023AWQ: Activation-aware Weight Quantization for LLM Compression and Accelerationcited by 1,234
- 2019Surrogate Gradient Learning in Spiking Neural Networkscited by 604
- 2020Pruning neural networks without any data by iteratively conserving synaptic flowcited by 475
- 2019HAWQ: Hessian AWare Quantization of Neural Networks with Mixed-Precisioncited by 387
- 2024The Era of 1-bit LLMs: All Large Language Models are in 1.58 Bitscited by 289
- 2021Training Spiking Neural Networks Using Lessons From Deep Learningcited by 251
- 2016Trained Ternary Quantizationcited by 203
- 2023BitNet: Scaling 1-bit Transformers for Large Language Modelscited by 175
- 2017Deep Rewiring: Training very sparse deep networkscited by 136
- 2016Convolutional Networks for Fast, Energy-Efficient Neuromorphic Computingcited by 68
12 more
- 2018Fisher Information and Natural Gradient Learning of Random Deep Networkscited by 28
- 2023Darwin3: A large-scale neuromorphic chip with a Novel ISA and On-Chip Learningcited by 22
- 2020End to End Binarized Neural Networks for Text Classificationcited by 13
- 2025QP-SNN: Quantized and Pruned Spiking Neural Networkscited by 12
- 2024Towards Efficient Deep Spiking Neural Networks Construction with Spiking Activity based Pruningcited by 9
- 2022The fine line between dead neurons and sparsity in binarized spiking neural networkscited by 7
- 2025Towards Superior Quantization Accuracy: A Layer-sensitive Approachcited by 7
- 2021Hessian Aware Quantization of Spiking Neural Networkscited by 5
- 2023Low Precision Quantization-aware Training in Spiking Neural Networks with Differentiable Quantization Functioncited by 3
- 2020A Power-Efficient Binary-Weight Spiking Neural Network Architecture for Real-Time Object Classificationcited by 1
- 2020Rethinking Clustering for Robustnesscited by 1
- 2021Prominent characteristics of recurrent neuronal networks are robust against low synaptic weight resolutioncited by 1
Built on by
Citing papers per year: 2026: 2.
2 papers.