A High Energy-Efficiency Multi-core Neuromorphic Architecture for Deep SNN Training

There is a growing necessity for edge training to adapt to dynamically changing environment. Neuromorphic computing represents a significant pathway for high-efficiency intelligent computation in energy-constrained edges, but existing neuromorphic architectures lack the ability of directly training spiking neural networks (SNNs) based on backpropagation. We develop a multi-core neuromorphic architecture with Feedforward-Propagation, Back-Propagation, and Weight-Gradient engines in each core, supporting high efficient parallel computing at both the engine and core levels. It combines various data flows and sparse computation optimization by fully leveraging the sparsity in SNN training, obtaining a high energy efficiency of 1.05TFLOPS/W@ FP16 @ 28nm, 55 ~ 85% reduction of DRAM access compared to A100 GPU in SNN trainings, and a 20-core deep SNN training and a 5-worker federated learning on FPGAs. Our study develops the first multi-core neuromorphic architecture supporting the direct SNN training, facilitating the neuromorphic computing in edge-learnable applications.

Flexible and Scalable Deep Dendritic Spiking Neural Networks with Multiple Nonlinear Branching

Recent advances in spiking neural networks (SNNs) have a predominant focus on network architectures, while relatively little attention has been paid to the underlying neuron model. The point neuron models, a cornerstone of deep SNNs, pose a bottleneck on the network-level expressivity since they depict somatic dynamics only. In contrast, the multi-compartment models in neuroscience offer remarkable expressivity by introducing dendritic morphology and dynamics, but remain underexplored in deep learning due to their unaffordable computational cost and inflexibility. To combine the advantages of both sides for a flexible, efficient yet more powerful model, we propose the dendritic spiking neuron (DendSN) incorporating multiple dendritic branches with nonlinear dynamics. Compared to the point spiking neurons, DendSN exhibits significantly higher expressivity. DendSN's flexibility enables its seamless integration into diverse deep SNN architectures. To accelerate dendritic SNNs (DendSNNs), we parallelize dendritic state updates across time steps, and develop Triton kernels for GPU-level acceleration. As a result, we can construct large-scale DendSNNs with depth comparable to their point SNN counterparts. Next, we comprehensively evaluate DendSNNs' performance on various demanding tasks. By modulating dendritic branch strengths using a context signal, catastrophic forgetting of DendSNNs is substantially mitigated. Moreover, DendSNNs demonstrate enhanced robustness against noise and adversarial attacks compared to point SNNs, and excel in few-shot learning settings. Our work firstly demonstrates the possibility of training bio-plausible dendritic SNNs with depths and scales comparable to traditional point SNNs, and reveals superior expressivity and robustness of reduced dendritic neuron models in deep learning, thereby offering a fresh perspective on advancing neural network design.

Scaling Spike-driven Transformer with Efficient Spike Firing Approximation Training

The ambition of brain-inspired Spiking Neural Networks (SNNs) is to become a low-power alternative to traditional Artificial Neural Networks (ANNs). This work addresses two major challenges in realizing this vision: the performance gap between SNNs and ANNs, and the high training costs of SNNs. We identify intrinsic flaws in spiking neurons caused by binary firing mechanisms and propose a Spike Firing Approximation (SFA) method using integer training and spike-driven inference. This optimizes the spike firing pattern of spiking neurons, enhancing efficient training, reducing power consumption, improving performance, enabling easier scaling, and better utilizing neuromorphic chips. We also develop an efficient spike-driven Transformer architecture and a spike-masked autoencoder to prevent performance degradation during SNN scaling. On ImageNet-1k, we achieve state-of-the-art top-1 accuracy of 78.5\%, 79.8\%, 84.0\%, and 86.2\% with models containing 10M, 19M, 83M, and 173M parameters, respectively. For instance, the 10M model outperforms the best existing SNN by 7.2\% on ImageNet, with training time acceleration and inference energy efficiency improved by 4.5$\times$ and 3.9$\times$, respectively. We validate the effectiveness and efficiency of the proposed method across various tasks, including object detection, semantic segmentation, and neuromorphic vision tasks. This work enables SNNs to match ANN performance while maintaining the low-power advantage, marking a significant step towards SNNs as a general visual backbone. Code is available at https://github.com/BICLab/Spike-Driven-Transformer-V3.

MetaLA: Unified Optimal Linear Approximation to Softmax Attention Map

Various linear complexity models, such as Linear Transformer (LinFormer), State Space Model (SSM), and Linear RNN (LinRNN), have been proposed to replace the conventional softmax attention in Transformer structures. However, the optimal design of these linear models is still an open question. In this work, we attempt to answer this question by finding the best linear approximation to softmax attention from a theoretical perspective. We start by unifying existing linear complexity models as the linear attention form and then identify three conditions for the optimal linear attention design: 1) Dynamic memory ability; 2) Static approximation ability; 3) Least parameter approximation. We find that none of the current linear models meet all three conditions, resulting in suboptimal performance. Instead, we propose Meta Linear Attention (MetaLA) as a solution that satisfies these conditions. Our experiments on Multi-Query Associative Recall (MQAR) task, language modeling, image classification, and Long-Range Arena (LRA) benchmark demonstrate that MetaLA is more effective than the existing linear models.

Towards Unifying Understanding and Generation in the Era of Vision Foundation Models: A Survey from the Autoregression Perspective

Autoregression in large language models (LLMs) has shown impressive scalability by unifying all language tasks into the next token prediction paradigm. Recently, there is a growing interest in extending this success to vision foundation models. In this survey, we review the recent advances and discuss future directions for autoregressive vision foundation models. First, we present the trend for next generation of vision foundation models, i.e., unifying both understanding and generation in vision tasks. We then analyze the limitations of existing vision foundation models, and present a formal definition of autoregression with its advantages. Later, we categorize autoregressive vision foundation models from their vision tokenizers and autoregression backbones. Finally, we discuss several promising research challenges and directions. To the best of our knowledge, this is the first survey to comprehensively summarize autoregressive vision foundation models under the trend of unifying understanding and generation. A collection of related resources is available at https://github.com/EmmaSRH/ARVFM.

Correlation-Aware Select and Merge Attention for Efficient Fine-Tuning and Context Length Extension

Modeling long sequences is crucial for various large-scale models; however, extending existing architectures to handle longer sequences presents significant technical and resource challenges. In this paper, we propose an efficient and flexible attention architecture that enables the extension of context lengths in large language models with reduced computational resources and fine-tuning time compared to other excellent methods. Specifically, we introduce correlation-aware selection and merging mechanisms to facilitate efficient sparse attention. In addition, we also propose a novel data augmentation technique involving positional encodings to enhance generalization to unseen positions. The results are as follows: First, using a single A100, we achieve fine-tuning on Llama2-7B with a sequence length of 32K, which is more efficient than other methods that rely on subsets for regression. Second, we present a comprehensive method for extending context lengths across the pre-training, fine-tuning, and inference phases. During pre-training, our attention mechanism partially breaks translation invariance during token selection, so we apply positional encodings only to the selected tokens. This approach achieves relatively high performance and significant extrapolation capabilities. For fine-tuning, we introduce Cyclic, Randomly Truncated, and Dynamically Growing NTK Positional Embedding (CRD NTK). This design allows fine-tuning with a sequence length of only 16K, enabling models such as Llama2-7B and Mistral-7B to perform inference with context lengths of up to 1M or even arbitrary lengths. Our method achieves 100\% accuracy on the passkey task with a context length of 4M and maintains stable perplexity at a 1M context length. This represents at least a 64-fold reduction in resource requirements compared to traditional full-attention mechanisms, while still achieving competitive performance.

Distance-Forward Learning: Enhancing the Forward-Forward Algorithm Towards High-Performance On-Chip Learning

The Forward-Forward (FF) algorithm was recently proposed as a local learning method to address the limitations of backpropagation (BP), offering biological plausibility along with memory-efficient and highly parallelized computational benefits. However, it suffers from suboptimal performance and poor generalization, largely due to inadequate theoretical support and a lack of effective learning strategies. In this work, we reformulate FF using distance metric learning and propose a distance-forward algorithm (DF) to improve FF performance in supervised vision tasks while preserving its local computational properties, making it competitive for efficient on-chip learning. To achieve this, we reinterpret FF through the lens of centroid-based metric learning and develop a goodness-based N-pair margin loss to facilitate the learning of discriminative features. Furthermore, we integrate layer-collaboration local update strategies to reduce information loss caused by greedy local parameter updates. Our method surpasses existing FF models and other advanced local learning approaches, with accuracies of 99.7\% on MNIST, 88.2\% on CIFAR-10, 59\% on CIFAR-100, 95.9\% on SVHN, and 82.5\% on ImageNette, respectively. Moreover, it achieves comparable performance with less than 40\% memory cost compared to BP training, while exhibiting stronger robustness to multiple types of hardware-related noise, demonstrating its potential for online learning and energy-efficient computation on neuromorphic chips.

Scalable Autoregressive Image Generation with Mamba

We introduce AiM, an autoregressive (AR) image generative model based on Mamba architecture. AiM employs Mamba, a novel state-space model characterized by its exceptional performance for long-sequence modeling with linear time complexity, to supplant the commonly utilized Transformers in AR image generation models, aiming to achieve both superior generation quality and enhanced inference speed. Unlike existing methods that adapt Mamba to handle two-dimensional signals via multi-directional scan, AiM directly utilizes the next-token prediction paradigm for autoregressive image generation. This approach circumvents the need for extensive modifications to enable Mamba to learn 2D spatial representations. By implementing straightforward yet strategically targeted modifications for visual generative tasks, we preserve Mamba's core structure, fully exploiting its efficient long-sequence modeling capabilities and scalability. We provide AiM models in various scales, with parameter counts ranging from 148M to 1.3B. On the ImageNet1K 256*256 benchmark, our best AiM model achieves a FID of 2.21, surpassing all existing AR models of comparable parameter counts and demonstrating significant competitiveness against diffusion models, with 2 to 10 times faster inference speed. Code is available at https://github.com/hp-l33/AiM

Integer-Valued Training and Spike-Driven Inference Spiking Neural Network for High-performance and Energy-efficient Object Detection

Brain-inspired Spiking Neural Networks (SNNs) have bio-plausibility and low-power advantages over Artificial Neural Networks (ANNs). Applications of SNNs are currently limited to simple classification tasks because of their poor performance. In this work, we focus on bridging the performance gap between ANNs and SNNs on object detection. Our design revolves around network architecture and spiking neuron. First, the overly complex module design causes spike degradation when the YOLO series is converted to the corresponding spiking version. We design a SpikeYOLO architecture to solve this problem by simplifying the vanilla YOLO and incorporating meta SNN blocks. Second, object detection is more sensitive to quantization errors in the conversion of membrane potentials into binary spikes by spiking neurons. To address this challenge, we design a new spiking neuron that activates Integer values during training while maintaining spike-driven by extending virtual timesteps during inference. The proposed method is validated on both static and neuromorphic object detection datasets. On the static COCO dataset, we obtain 66.2% mAP@50 and 48.9% mAP@50:95, which is +15.0% and +18.7% higher than the prior state-of-the-art SNN, respectively. On the neuromorphic Gen1 dataset, we achieve 67.2% mAP@50, which is +2.5% greater than the ANN with equivalent architecture, and the energy efficiency is improved by 5.7*. Code: https://github.com/BICLab/SpikeYOLO

Unveiling the Potential of Spiking Dynamics in Graph Representation Learning through Spatial-Temporal Normalization and Coding Strategies

In recent years, spiking neural networks (SNNs) have attracted substantial interest due to their potential to replicate the energy-efficient and event-driven processing of biological neurons. Despite this, the application of SNNs in graph representation learning, particularly for non-Euclidean data, remains underexplored, and the influence of spiking dynamics on graph learning is not yet fully understood. This work seeks to address these gaps by examining the unique properties and benefits of spiking dynamics in enhancing graph representation learning. We propose a spike-based graph neural network model that incorporates spiking dynamics, enhanced by a novel spatial-temporal feature normalization (STFN) technique, to improve training efficiency and model stability. Our detailed analysis explores the impact of rate coding and temporal coding on SNN performance, offering new insights into their advantages for deep graph networks and addressing challenges such as the oversmoothing problem. Experimental results demonstrate that our SNN models can achieve competitive performance with state-of-the-art graph neural networks (GNNs) while considerably reducing computational costs, highlighting the potential of SNNs for efficient neuromorphic computing applications in complex graph-based scenarios.