arXiv subjects

Lei Deng

Publications and source records attributed to Lei Deng.

At least 37 records · Page 2Linked to original sources

arXiv2025

Understanding the Functional Roles of Modelling Components in Spiking Neural Networks

Spiking neural networks (SNNs), inspired by the neural circuits of the brain, are promising in achieving high computational efficiency with biological fidelity. Nevertheless, it is quite difficult to optimize SNNs because the functional roles of their modelling components remain unclear. By designing and evaluating several variants of the classic model, we systematically investigate the functional roles of key modelling components, leakage, reset, and recurrence, in leaky integrate-and-fire (LIF) based SNNs. Through extensive experiments, we demonstrate how these components influence the accuracy, generalization, and robustness of SNNs. Specifically, we find that the leakage plays a crucial role in balancing memory retention and robustness, the reset mechanism is essential for uninterrupted temporal processing and computational efficiency, and the recurrence enriches the capability to model complex dynamics at a cost of robustness degradation. With these interesting observations, we provide optimization suggestions for enhancing the performance of SNNs in different scenarios. This work deepens the understanding of how SNNs work, which offers valuable guidance for the development of more effective and robust neuromorphic models.

cs.NE↗

arXiv2024

Beyond Scaling Laws: Understanding Transformer Performance with Associative Memory

Increasing the size of a Transformer does not always lead to enhanced performance. This phenomenon cannot be explained by the empirical scaling laws. Furthermore, the model's enhanced performance is closely associated with its memorization of the training samples. We present a theoretical framework that sheds light on the memorization during pre-training of transformer-based language models. We model the behavior of Transformers with associative memories using Hopfield networks, such that each transformer block effectively conducts an approximate nearest-neighbor search. In particular, the energy function in modern continuous Hopfield networks serves as an explanation for the attention mechanism, which we approximate with a distance-based energy function. By observing that the softmax function corresponds to the gradient of the LogSumExp function in the energy, and employing the majorization-minimization technique, we construct a global energy function designed to capture the layered architecture. We demonstrate a dependency between the model size and the dataset size for the model to achieve optimal performance, and we show that the achievable cross-entropy loss is bounded from below.

cs.LG↗

arXiv2024

High Perceptual Quality Wireless Image Delivery with Denoising Diffusion Models

We consider the image transmission problem over a noisy wireless channel via deep learning-based joint source-channel coding (DeepJSCC) along with a denoising diffusion probabilistic model (DDPM) at the receiver. Specifically, we are interested in the perception-distortion trade-off in the practical finite block length regime, in which separate source and channel coding can be highly suboptimal. We introduce a novel scheme, where the conventional DeepJSCC encoder targets transmitting a lower resolution version of the image, which later can be refined thanks to the generative model available at the receiver. In particular, we utilize the range-null space decomposition of the target image; DeepJSCC transmits the range-space of the image, while DDPM progressively refines its null space contents. Through extensive experiments, we demonstrate significant improvements in distortion and perceptual quality of reconstructed images compared to standard DeepJSCC and the state-of-the-art generative learning-based method.

eess.IV↗

arXiv2024

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.

cs.NE↗

arXiv2024

A Mean Field Ansatz for Zero-Shot Weight Transfer

The pre-training cost of large language models (LLMs) is prohibitive. One cutting-edge approach to reduce the cost is zero-shot weight transfer, also known as model growth for some cases, which magically transfers the weights trained in a small model to a large model. However, there are still some theoretical mysteries behind the weight transfer. In this paper, inspired by prior applications of mean field theory to neural network dynamics, we introduce a mean field ansatz to provide a theoretical explanation for weight transfer. Specifically, we propose the row-column (RC) ansatz under the mean field point of view, which describes the measure structure of the weights in the neural network (NN) and admits a close measure dynamic. Thus, the weights of different sizes NN admit a common distribution under proper assumptions, and weight transfer methods can be viewed as sampling methods. We empirically validate the RC ansatz by exploring simple MLP examples and LLMs such as GPT-3 and Llama-3.1. We show the mean-field point of view is adequate under suitable assumptions which can provide theoretical support for zero-shot weight transfer.

cs.LG↗

arXiv2024

Closeby Habitable Exoplanet Survey (CHES). II. An Observation Strategy for the Target Stars

The Closeby Habitable Exoplanet Survey (CHES) constitutes a mission intricately designed to systematically survey approximately 100 solar-type stars located within the immediate proximity of the solar system, specifically within a range of 10 parsecs. The core objective of this mission is the detection and characterization of potentially habitable Earth-like planets or super-Earths within the habitable zone of these stars. The CHES mission obtains high-precision astrometric measurements of planets orbiting the target stars by observing angular distance variations between the target star and reference stars. As a result, we surveyed the relevant parameters of both target and reference stars in detail, conducting a thorough analysis and calculation of the required observation accuracy, the number of observations, and the priority assigned to each target star. Observational emphasis will be concentrated on targets considered of higher priority, ensuring the effectiveness of their observation capabilities. Through this approach, we formulate a five-year observation strategy that will cover all the target stars within a six-month timeframe. The strategy not only fulfills the required observing capability but also exhibit high efficiency simultaneously, providing an executable program for future mission. Over the span of the mission's five-year duration, a cumulative observation time of 29,220 hours will be available. Approximately 86 percent of this, totaling 25,120 hours, is allocated for the observation of target stars. This allocation leaves approximately 4,100 hours for extended scientific observation programs. We have also performed simulated observations based on this strategy and verified its observational capability for exoplanets.

astro-ph.EP↗

arXiv2024

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.

cs.AI↗

arXiv2024

Retrieval Meets Reasoning: Dynamic In-Context Editing for Long-Text Understanding

Current Large Language Models (LLMs) face inherent limitations due to their pre-defined context lengths, which impede their capacity for multi-hop reasoning within extensive textual contexts. While existing techniques like Retrieval-Augmented Generation (RAG) have attempted to bridge this gap by sourcing external information, they fall short when direct answers are not readily available. We introduce a novel approach that re-imagines information retrieval through dynamic in-context editing, inspired by recent breakthroughs in knowledge editing. By treating lengthy contexts as malleable external knowledge, our method interactively gathers and integrates relevant information, thereby enabling LLMs to perform sophisticated reasoning steps. Experimental results demonstrate that our method effectively empowers context-limited LLMs, such as Llama2, to engage in multi-hop reasoning with improved performance, which outperforms state-of-the-art context window extrapolation methods and even compares favorably to more advanced commercial long-context models. Our interactive method not only enhances reasoning capabilities but also mitigates the associated training and computational costs, making it a pragmatic solution for enhancing LLMs' reasoning within expansive contexts.

cs.CL↗

arXiv2024

Enhancing Graph Representation Learning with Attention-Driven Spiking Neural Networks

Graph representation learning has become a crucial task in machine learning and data mining due to its potential for modeling complex structures such as social networks, chemical compounds, and biological systems. Spiking neural networks (SNNs) have recently emerged as a promising alternative to traditional neural networks for graph learning tasks, benefiting from their ability to efficiently encode and process temporal and spatial information. In this paper, we propose a novel approach that integrates attention mechanisms with SNNs to improve graph representation learning. Specifically, we introduce an attention mechanism for SNN that can selectively focus on important nodes and corresponding features in a graph during the learning process. We evaluate our proposed method on several benchmark datasets and show that it achieves comparable performance compared to existing graph learning techniques.

cs.AI↗

arXiv2024

Extreme Video Compression with Pre-trained Diffusion Models

Diffusion models have achieved remarkable success in generating high quality image and video data. More recently, they have also been used for image compression with high perceptual quality. In this paper, we present a novel approach to extreme video compression leveraging the predictive power of diffusion-based generative models at the decoder. The conditional diffusion model takes several neural compressed frames and generates subsequent frames. When the reconstruction quality drops below the desired level, new frames are encoded to restart prediction. The entire video is sequentially encoded to achieve a visually pleasing reconstruction, considering perceptual quality metrics such as the learned perceptual image patch similarity (LPIPS) and the Frechet video distance (FVD), at bit rates as low as 0.02 bits per pixel (bpp). Experimental results demonstrate the effectiveness of the proposed scheme compared to standard codecs such as H.264 and H.265 in the low bpp regime. The results showcase the potential of exploiting the temporal relations in video data using generative models. Code is available at: https://github.com/ElesionKyrie/Extreme-Video-Compression-With-Prediction-Using-Pre-trainded-Diffusion-Models-

eess.IV↗

arXiv2023

Extending Context Window of Large Language Models via Semantic Compression

Transformer-based Large Language Models (LLMs) often impose limitations on the length of the text input to ensure the generation of fluent and relevant responses. This constraint restricts their applicability in scenarios involving long texts. We propose a novel semantic compression method that enables generalization to texts that are 6-8 times longer, without incurring significant computational costs or requiring fine-tuning. Our proposed framework draws inspiration from source coding in information theory and employs a pre-trained model to reduce the semantic redundancy of long inputs before passing them to the LLMs for downstream tasks. Experimental results demonstrate that our method effectively extends the context window of LLMs across a range of tasks including question answering, summarization, few-shot learning, and information retrieval. Furthermore, the proposed semantic compression method exhibits consistent fluency in text generation while reducing the associated computational overhead.

cs.CL↗

arXiv2023

NP-Hardness of Tensor Network Contraction Ordering

We study the optimal order (or sequence) of contracting a tensor network with a minimal computational cost. We conclude 2 different versions of this optimal sequence: that minimize the operation number (OMS) and that minimize the time complexity (CMS). Existing results only shows that OMS is NP-hard, but no conclusion on CMS problem. In this work, we firstly reduce CMS to CMS-0, which is a sub-problem of CMS with no free indices. Then we prove that CMS is easier than OMS, both in general and in tree cases. Last but not least, we prove that CMS is still NP-hard. Based on our results, we have built up relationships of hardness of different tensor network contraction problems.

cs.CC↗

arXiv2023

An Input-Queueing TSN Switching Architecture to Achieve Zero Packet Loss for Timely Traffic

Zero packet loss with bounded latency is necessary for many applications, such as industrial control networks, automotive Ethernet, and aircraft communication systems. Traditional networks cannot meet the such strict requirement, and thus Time-Sensitive Networking (TSN) emerges. TSN is a set of standards proposed by IEEE 802 for providing deterministic connectivity in terms of low packet loss, low packet delay variation, and guaranteed packet transport. However, to our knowledge, few existing TSN solutions can deterministically achieve zero packet loss with bounded latency. This paper fills in this blank by proposing a novel input-queueing TSN switching architecture, under which we design a TDMA-like scheduling policy (called M-TDMA) along with a sufficient condition and an EDF-like scheduling policy (called M-EDF) along with a different sufficient condition to achieve zero packet loss with bounded latency.

cs.NI↗

arXiv2023

Advancing Spiking Neural Networks towards Deep Residual Learning

Despite the rapid progress of neuromorphic computing, inadequate capacity and insufficient representation power of spiking neural networks (SNNs) severely restrict their application scope in practice. Residual learning and shortcuts have been evidenced as an important approach for training deep neural networks, but rarely did previous work assess their applicability to the characteristics of spike-based communication and spatiotemporal dynamics. In this paper, we first identify that this negligence leads to impeded information flow and the accompanying degradation problem in previous residual SNNs. To address this issue, we propose a novel SNN-oriented residual architecture termed MS-ResNet, which establishes membrane-based shortcut pathways, and further prove that the gradient norm equality can be achieved in MS-ResNet by introducing block dynamical isometry theory, which ensures the network can be well-behaved in a depth-insensitive way. Thus we are able to significantly extend the depth of directly trained SNNs, e.g., up to 482 layers on CIFAR-10 and 104 layers on ImageNet, without observing any slight degradation problem. To validate the effectiveness of MS-ResNet, experiments on both frame-based and neuromorphic datasets are conducted. MS-ResNet104 achieves a superior result of 76.02% accuracy on ImageNet, which is the highest to our best knowledge in the domain of directly trained SNNs. Great energy efficiency is also observed, with an average of only one spike per neuron needed to classify an input sample. We believe our powerful and scalable models will provide a strong support for further exploration of SNNs.

cs.NE↗

arXiv2022

Solar Ring Mission: Building a Panorama of the Sun and Inner-heliosphere

Solar Ring (SOR) is a proposed space science mission to monitor and study the Sun and inner heliosphere from a full 360° perspective in the ecliptic plane. It will deploy three 120°-separated spacecraft on the 1-AU orbit. The first spacecraft, S1, locates 30° upstream of the Earth, the second, S2, 90° downstream, and the third, S3, completes the configuration. This design with necessary science instruments, e.g., the Doppler-velocity and vector magnetic field imager, wide-angle coronagraph, and in-situ instruments, will allow us to establish many unprecedented capabilities: (1) provide simultaneous Doppler-velocity observations of the whole solar surface to understand the deep interior, (2) provide vector magnetograms of the whole photosphere - the inner boundary of the solar atmosphere and heliosphere, (3) provide the information of the whole lifetime evolution of solar featured structures, and (4) provide the whole view of solar transients and space weather in the inner heliosphere. With these capabilities, Solar Ring mission aims to address outstanding questions about the origin of solar cycle, the origin of solar eruptions and the origin of extreme space weather events. The successful accomplishment of the mission will construct a panorama of the Sun and inner-heliosphere, and therefore advance our understanding of the star and the space environment that holds our life.

astro-ph.SR↗

arXiv2022

Attention Spiking Neural Networks

Benefiting from the event-driven and sparse spiking characteristics of the brain, spiking neural networks (SNNs) are becoming an energy-efficient alternative to artificial neural networks (ANNs). However, the performance gap between SNNs and ANNs has been a great hindrance to deploying SNNs ubiquitously for a long time. To leverage the full potential of SNNs, we study the effect of attention mechanisms in SNNs. We first present our idea of attention with a plug-and-play kit, termed the Multi-dimensional Attention (MA). Then, a new attention SNN architecture with end-to-end training called "MA-SNN" is proposed, which infers attention weights along the temporal, channel, as well as spatial dimensions separately or simultaneously. Based on the existing neuroscience theories, we exploit the attention weights to optimize membrane potentials, which in turn regulate the spiking response in a data-dependent way. At the cost of negligible additional parameters, MA facilitates vanilla SNNs to achieve sparser spiking activity, better performance, and energy efficiency concurrently. Experiments are conducted in event-based DVS128 Gesture/Gait action recognition and ImageNet-1k image classification. On Gesture/Gait, the spike counts are reduced by 84.9%/81.6%, and the task accuracy and energy efficiency are improved by 5.9%/4.7% and 3.4$\times$/3.2$\times$. On ImageNet-1K, we achieve top-1 accuracy of 75.92% and 77.08% on single/4-step Res-SNN-104, which are state-of-the-art results in SNNs. To our best knowledge, this is for the first time, that the SNN community achieves comparable or even better performance compared with its ANN counterpart in the large-scale dataset. Our work lights up SNN's potential as a general backbone to support various applications for SNNs, with a great balance between effectiveness and efficiency.

cs.CV↗

arXiv2022

The Story of $1/e$: ALOHA-based and Reinforcement-Learning-based Random Access for Delay-Constrained Communications

Motivated by the proliferation of real-time applications in multimedia communication systems, tactile Internet, and cyber-physical systems, supporting delay-constrained traffic becomes critical for such systems. In delay-constrained traffic, each packet has a hard deadline; when it is not delivered before its deadline is up, it becomes useless and will be removed from the system. In this work, we focus on designing random access schemes for delay-constrained wireless communications. We first investigate three ALOHA-based schemes and prove that the system timely throughput of all three schemes under corresponding optimal transmission probabilities asymptotically converges to $1/e$, same as the well-known throughput limit for delay-unconstrained ALOHA systems. The fundamental reason why ALOHA-based schemes cannot achieve asymptotical system timely throughput beyond $1/e$ is that all active ALOHA stations access the channel with the same probability in any slot. To go beyond $1/e$, we propose a reinforcement-learning-based scheme for delay-constrained wireless communications, called RLRA-DC, under which different stations collaboratively attain different transmission probabilities by only interacting with the access point. Our numerical result shows that the system timely throughput of RLRA-DC can be as high as 0.8 for tens of stations and can still reach 0.6 even for thousands of stations, much larger than $1/e$.

cs.IT↗

arXiv2022

CHES: a space-borne astrometric mission for the detection of habitable planets of the nearby solar-type stars

The Closeby Habitable Exoplanet Survey (CHES) mission is proposed to discover habitable-zone Earth-like planets of the nearby solar-type stars ($\sim 10~\mathrm{pc}$ away from our solar system) via micro-arcsecond relative astrometry. The major scientific objectives of CHES are: to search for Earth Twins or terrestrial planets in habitable zones orbiting 100 FGK nearby stars; further to conduct a comprehensive survey and extensively characterize the nearby planetary systems. The primary payload is a high-quality, low-distortion, high-stability telescope. The optical subsystem is a coaxial three-mirror anastigmat (TMA) with a $1.2 \mathrm{~m}$-aperture, $0.44^{\circ} \times 0.44^{\circ}$ field of view and $500 \mathrm{~nm}-900 \mathrm{~nm}$ working waveband. The camera focal plane is composed of 81 MOSAIC scientific CMOS detectors each with $4 \mathrm{~K} \times 4 \mathrm{~K}$ pixels. The heterodyne laser interferometric calibration technology is employed to ensure micro-arcsecond level (1 $μ$as) relative astrometry precision to meet the requirements for detection of Earth-like planets. CHES satellite operates at the Sun-Earth L2 point and observes the entire target stars for 5 years. CHES will offer the first direct measurements of true masses and inclinations of Earth Twins and super-Earths orbiting our neighbor stars based on micro-arcsecond astrometry from space. This will definitely enhance our understanding of the formation of diverse nearby planetary systems and the emergence of other worlds for solar-type stars, and finally to reflect the evolution of our own solar system.

astro-ph.EP↗