Organizations: Institute for Artificial Intelligence, Peking University, and also with Beijing Key Laboratory of Brain-inspired Spiking Large Models, the School of Computer Science, Peking University, Beijing, China · School of Computer Science, Peking University, and also with the National Key Laboratory for Multimedia Information Processing, Peking University, Beijing, China
Spiking Neural Networks (SNNs) serve as core architectures for neuromorphic computing thanks to event-driven operation and ultra-low power consumption. Direct SNN training is hindered by non-differentiable spikes that induce vanishing gradients and unstable optimization. ANN-SNN conversion circumvents such issues by reusing well-trained ANN weights for low-latency, energy-efficient inference. Nevertheless, existing conversion schemes suffer from severe accuracy drops at small timesteps, large inference delays and cumulative quantization errors, even with marginal performance loss at large T. To address these limitations, we first analyze flaws of conventional conversion pipelines from residual membrane potential statistics and propose a novel conversion strategy combining dynamic initial potential tuning and feature enhancement. We then introduce a regularization loss LRMPD to adapt initial potential of IF neurons and mitigate systematic truncation bias from boundary aggregation. A dedicated SCR-Conv2d competitive refinement layer with grouped convolution is further built to sharpen feature discrimination, eliminate redundant spikes and stabilize encoding under tiny time windows. Integrated with the state-of-the-art QCFS baseline, our approach delivers consistent low-latency performance gains and generalizes to ReLU CNNs, ANN Transformers, and multi-threshold SNN variants. Evaluations on CIFAR-10, CIFAR-100 and ImageNet verify prominent accuracy improvements at T=2,4,8, with negligible extra computation overhead. This work offers an effective conversion paradigm to facilitate real-world SNN deployment on neuromorphic chips.
Figures & tables
Fig. 1: Distribution of residual membrane potential. The subplots demonstrate RMPD under distinct input distributions. Case (a) is dominated by the interval [Lθ,L2θ] , while case (b) receives equal contributions from [0,Lθ] and [Lθ,L2θ] . Correspondingly, the RMP distribution in (a) exhibits near-Gaussian behavior, whereas the distribution in (b) severely departs from Gaussianity. Figure (c),(d) correspond to figure (a),(b) with smaller variance. Comparing case (c),(d) with (a),(b), we can tell that reducing variance of neuron input amplify the error-reduction effect.
Fig. 2: Fine-tuning of the initial membrane potential minit . The gradient step for updating minit is proportional to the offset of the target interval. After each adjustment, the resulting RMPD becomes closer to a normal distribution, and the corresponding spike activations cluster more tightly around the characteristic firing frequency of each neuron.
Fig. 3: RMPD shapes before and after adjustment. Parameter QQ quantifies the similarity to Gaussian distribution, and it can be observed that the regularized RMPD is closer to Gaussian distribution after adjustment.
Fig. 4: Fine-tuning of initial membrane potential with shift compensation
Fig. 5: Architecture of SCR-Conv2d. The SCR-Conv2d layer is inserted after the standard convolutional layer. The outputs of the IF layer are transformed into refinement signal magnitudes via the nonlinear function Fnl . An additional IF layer is appended after SCR-Conv2d to maintain spike transmission. This figure illustrates the spike-driven information propagation pipeline of the converted SNN within VGG-16.
Method
Net.
Param. (M)
ANN
2
4
8
16
32
64
CIFAR-10
QCFS [ 41 ]
VGG16
138.36
95.78
90.67
94.21
95.40
95.73
95.74
95.83
SNM [ 44 ]
VGG16
138.36
95.72
90.55
94.54
95.56
95.78
95.77
95.82
SRP( τ =4) [ 18 ]
VGG16
138.36
95.62
−
95.43
95.55
95.58
95.57
95.60
NQQCFS [ 37 ]
VGG16
138.36
95.21
91.93
94.80
95.48
95.70
95.79
95.72
SCR-SRMP
VGG16
138.36
95.65
92.84
94.80
95.66
95.78
95.80
95.86
TABLE I: Comparison with state-of-the-art methods on the CIFAR-10,CIFAR-100 and ImageNet dataset
Arch.
Method
Param. (M)
Original (ANN)
SNN
T=2
T=4
CIFAR-10
ViT-S/16
ECMT [ 69 ]
22
98.33
24.81
92.78
ViT-S/16
SRMP
22
98.33
26.62
94.31
ViT-B/16
ECMT [ 69 ]
86
98.73
85.67
96.76
ViT-B/16
SRMP
86
98.73
87.17
97.03
TABLE II: Comparison results on Transformer
Method
Conversion Type
Arch.
Param. (M)
T
Tratio
Acc(%)
QCFS [ 41 ]
ANN-to-SNN
ResNet-34
21.8
64
−
72.35
VGG-16
138
64
−
72.85
SRP [ 18 ]
ANN-to-SNN
ResNet-34
21.8
4(64)
0.0625
66.71(68.61)
VGG-16
138
4(64)
0.0625
66.46(69.43)
MST [ 73 ]
Transformer-to-SNN
Swin-T(BN)
28.5
128(512)
0.25
77.88(78.51)
STA [ 74 ]
Transformer-to-SNN
ViT-B/32
86
32(256)
0.125
78.72(82.79)
TABLE III: Comparison with previous works on ImageNet1k dataset
Fig. 6: Variations of Loss and LRMPD during fine-tuning on CIFAR-10, CIFAR-100 and ImageNet. LRMPD steadily decline, while Loss fluctuate mildly or rise slightly.
Fig. 7: Curves of LRMPD induction in ViT-S/16, ViT-B/16 and ViT-L/16 model on CIFAR-10, CIFAR-100, ImageNet. ALL the curves decline steadily and flatten out over time.
Method
Net.
ANN
2
4
8
16
CIFAR-10
QCFS [ 41 ]
VGG16
95.78
90.67
94.21
95.40
95.73
SRMP
VGG16
95.60
91.49
94.82
95.66
95.85
SCR
VGG16
95.65
91.92
94.36
95.40
95.77
SCR-SRMP
VGG16
95.65
92.84
94.80
95.46
95.78
QCFS [ 41 ]
ResNet-18
96.48
92.86
94.92
96.21
96.55
TABLE IV: Comparison of test results among SRMP, SCR-SRMP, SCR and QCFS on CIFAR-10
Fig. 8: Comparison with fixed weight and dynamic λp for LRMPD .
Fig. 9: Distribution of σ/θ in SNN layers in VGG16, ResNet18 and ResNet-20. Frequency of σ/θ<0.125 amounts to 84.67% in VGG-16, which indicates that at least 84.67% of neurons satisfy the conditions of Theorem 1 . Similar observations can be obtained for other models.
Spiking neural networks process information through discrete, time-dependent neural events, aiming for efficient computation and biologically inspired learning.
On neuromorphic chips, spiking neural networks (SNNs) consume far less energy than artificial neural networks (ANNs), but the limited memory of these chips requires low-bit synaptic weights and membrane potentials. Existing post-training quantization (PTQ) methods provide them without retraining and, like those for ANNs, minimize numerical errors such as the mean squared error (MSE). In an SNN, however, a quantization error reaches later layers only through the spikes it flips. We find that errors of equal MSE can cause very different accuracy losses, whereas errors that flip the same fraction of spikes cause the same loss. We therefore propose SpikeQ, a spike-guided PTQ framework. For weights, we select the level of each weight by simulating the downstream neurons and minimizing their spike errors. For membrane potentials, most existing methods quantize the state after the reset, whereas neuromorphic chips compare the threshold with the membrane potential stored at a fixed bit width. We target this pre-fire quantization and build a reset-aware grid that places the threshold on a rounding boundary, so rounding never flips the spike at the current time step. We evaluate SpikeQ on seven benchmarks of image classification, semantic segmentation, and object detection. With 2-bit pre-fire membrane potentials, every existing grid collapses, whereas SpikeQ retains most of the full-precision accuracy. With both at 4 bits, the models stay within 2.59 points of full precision, shrink their memory footprint 7.8--7.9 times, and use 6.4--6.9 times less energy.
Kai Sun, Peibo Duan, Zhanglu Yan +3
Monash University · National University of Singapore · Northeastern University
Most ANN-to-SNN conversion methods rely on a specific correspondence between the source activation and the spiking neuron dynamics. We propose a finite-state continuous-time Markov chain (CTMC) neuron framework whose stationary spike flux can approximate every continuous nonnegative monotone activation function on a compact interval. For a generalized CTMC family with affine input-dependent transitions, we prove uniform approximation to arbitrary accuracy over this function class and derive an explicit approximation error bound. In practice, two- and three-state CTMCs fit ReLU, sigmoid, softplus, and clipped ReLU on the evaluated input ranges, and we evaluate corresponding MLP conversions for each activation with layerwise rate scaling. Moderate clipping improves the conversion cost-accuracy tradeoff on the MNIST MLP and reduces SynOps by 27% on VGG-11/MNIST at matched ANN-SNN accuracy gap criteria, whereas the trend reverses on VGG-11/CIFAR-10. Mean-field and layerwise diagnostics indicate that finite-window sampling and terminal-layer mismatch are the main residual errors. Overall, our results establish finite-state CTMC neurons as a theoretically grounded framework for activation-flexible ANN-to-SNN conversion beyond fixed activation-neuron correspondences.
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