arXiv subjects

Bobin Deng

Publications and source records attributed to Bobin Deng.

9 recordsLinked to original sources

arXiv2026

NeuroPrefetcher: Storage-Aware Sparse LLM Inference via Delta Prefetching

Deploying large language models on edge devices is increasingly limited by a widening gap between model size and available memory. Existing approaches such as quantization, smaller models, and offloading can raise the effective memory limit, but they still assume that the model can be compressed or partitioned to fit within some budget. We target the harder model-exceeds-memory setting, in which the model remains larger than resident memory throughout execution and storage becomes an active source of weights on the critical path. We observe that MLP activity during autoregressive decoding has strong temporal locality: approximately 82-85% of active neurons persist from one token to the next. This means that most sparse weights needed for the current token are already resident, and only the newly needed rows must be fetched from storage. We present NeuroPrefetcher, a storage-backed LLM inference system that exploits this property through predictive delta prefetching. After layer 0, a single GPU-resident predictor, occupying 2.86% of base model parameters, predicts sparse activity for all downstream MLP layers in one forward pass. The runtime compares these predictions against resident GPU buffers and issues application-scheduled NVMe reads only for incoming delta rows, replacing reactive operating-system demand paging with explicit, model-aware weight movement. On real unified-memory edge hardware, NeuroPrefetcher achieves 7.9-12.0x speedup over llama.cpp across constrained memory budgets.

cs.DC↗

arXiv2026

FedMTFI: Feature Importance Based Optimized Multi Teacher Knowledge Distillation in Heterogeneous Federated Learning Environment

Federated learning (FL) is a decentralized approach that enables collaborative model training without exposing raw data. Instead of transferring sensitive data, it allows devices to share only model weights, keeping personal data locally and secure. However, in real world settings, the data held by devices is often not evenly distributed and devices mostly differ in computing power and memory capacity. These differences make FL harder to maintain consistent performance across the system. To address these issues, we propose FedMTFI, a novel architecture that combines multi-teacher knowledge distillation (MTKD) with feature importance to improve the FL process in heterogeneous environments. In FedMTFI, clients are clustered based on similar hardware and model types. Each cluster trains a specific model on not independently and identically distributed (non-IID) data. Within a cluster, every client updates that model using only its own local private data. The server then aggregates the locally trained models in each cluster using FedAvg to form multiple prototype models. Then these prototypes serve as teacher models to train a global generalized student model using MTKD. What makes FedMTFI more unique is the integration of Shapley values (SHAP) to emphasize important features during distillation, which enhances both accuracy and interpretability. Experimental results show that FedMTFI achieves higher accuracy than traditional FL algorithms and performs more effectively under non-IID data conditions.

cs.LG↗

arXiv2025

Characterizing and Understanding Energy Footprint and Efficiency of Small Language Model on Edges

Cloud-based large language models (LLMs) and their variants have significantly influenced real-world applications. Deploying smaller models (i.e., small language models (SLMs)) on edge devices offers additional advantages, such as reduced latency and independence from network connectivity. However, edge devices' limited computing resources and constrained energy budgets challenge efficient deployment. This study evaluates the power efficiency of five representative SLMs - Llama 3.2, Phi-3 Mini, TinyLlama, and Gemma 2 on Raspberry Pi 5, Jetson Nano, and Jetson Orin Nano (CPU and GPU configurations). Results show that Jetson Orin Nano with GPU acceleration achieves the highest energy-to-performance ratio, significantly outperforming CPU-based setups. Llama 3.2 provides the best balance of accuracy and power efficiency, while TinyLlama is well-suited for low-power environments at the cost of reduced accuracy. In contrast, Phi-3 Mini consumes the most energy despite its high accuracy. In addition, GPU acceleration, memory bandwidth, and model architecture are key in optimizing inference energy efficiency. Our empirical analysis offers practical insights for AI, smart systems, and mobile ad-hoc platforms to leverage tradeoffs from accuracy, inference latency, and power efficiency in energy-constrained environments.

cs.DC↗

arXiv2025

AgriSentinel: Privacy-Enhanced Embedded-LLM Crop Disease Alerting System

Crop diseases pose significant threats to global food security, agricultural productivity, and sustainable farming practices, directly affecting farmers' livelihoods and economic stability. To address the growing need for effective crop disease management, AI-based disease alerting systems have emerged as promising tools by providing early detection and actionable insights for timely intervention. However, existing systems often overlook critical aspects such as data privacy, market pricing power, and farmer-friendly usability, leaving farmers vulnerable to privacy breaches and economic exploitation. To bridge these gaps, we propose AgriSentinel, the first Privacy-Enhanced Embedded-LLM Crop Disease Alerting System. AgriSentinel incorporates a differential privacy mechanism to protect sensitive crop image data while maintaining classification accuracy. Its lightweight deep learning-based crop disease classification model is optimized for mobile devices, ensuring accessibility and usability for farmers. Additionally, the system includes a fine-tuned, on-device large language model (LLM) that leverages a curated knowledge pool to provide farmers with specific, actionable suggestions for managing crop diseases, going beyond simple alerting. Comprehensive experiments validate the effectiveness of AgriSentinel, demonstrating its ability to safeguard data privacy, maintain high classification performance, and deliver practical, actionable disease management strategies. AgriSentinel offers a robust, farmer-friendly solution for automating crop disease alerting and management, ultimately contributing to improved agricultural decision-making and enhanced crop productivity.

cs.CR↗

arXiv2025

A Sparsity Predicting Approach for Large Language Models via Activation Pattern Clustering

Large Language Models (LLMs) exhibit significant activation sparsity, where only a subset of neurons are active for a given input. Although this sparsity presents opportunities to reduce computational cost, efficiently utilizing it requires predicting activation patterns in a scalable manner. However, direct prediction at the neuron level is computationally expensive due to the vast number of neurons in modern LLMs. To enable efficient prediction and utilization of activation sparsity, we propose a clustering-based activation pattern compression framework. Instead of treating each neuron independently, we group similar activation patterns into a small set of representative clusters. Our method achieves up to 79.34% clustering precision, outperforming standard binary clustering approaches while maintaining minimal degradation in perplexity (PPL) scores. With a sufficiently large number of clusters, our approach attains a PPL score as low as 12.49, demonstrating its effectiveness in preserving model quality while reducing computational overhead. By predicting cluster assignments rather than individual neuron states, future models can efficiently infer activation patterns from pre-computed centroids. We detail the clustering algorithm, analyze its effectiveness in capturing meaningful activation structures, and demonstrate its potential to improve sparse computation efficiency. This clustering-based formulation serves as a foundation for future work on activation pattern prediction, paving the way for efficient inference in large-scale language models.

cs.LG↗

arXiv2025

The Robustness of Spiking Neural Networks in Federated Learning with Compression Against Non-omniscient Byzantine Attacks

Spiking Neural Networks (SNNs), which offer exceptional energy efficiency for inference, and Federated Learning (FL), which offers privacy-preserving distributed training, is a rising area of interest that highly beneficial towards Internet of Things (IoT) devices. Despite this, research that tackles Byzantine attacks and bandwidth limitation in FL-SNNs, both poses significant threats on model convergence and training times, still remains largely unexplored. Going beyond proposing a solution for both of these problems, in this work we highlight the dual benefits of FL-SNNs, against non-omniscient Byzantine adversaries (ones that restrict attackers access to local clients datasets), and greater communication efficiency, over FL-ANNs. Specifically, we discovered that a simple integration of Top-\k{appa} sparsification into the FL apparatus can help leverage the advantages of the SNN models in both greatly reducing bandwidth usage and significantly boosting the robustness of FL training against non-omniscient Byzantine adversaries. Most notably, we saw a massive improvement of roughly 40% accuracy gain in FL-SNNs training under the lethal MinMax attack

cs.CR↗

arXiv2024

Activation Sparsity Opportunities for Compressing General Large Language Models

Deploying local AI models, such as Large Language Models (LLMs), to edge devices can substantially enhance devices' independent capabilities, alleviate the server's burden, and lower the response time. Owing to these tremendous potentials, many big tech companies have released several lightweight Small Language Models (SLMs) to bridge this gap. However, we still have huge motivations to deploy more powerful (LLMs) AI models on edge devices and enhance their smartness level. Unlike the conventional approaches for AI model compression, we investigate activation sparsity. The activation sparsity method is orthogonal and combinable with existing techniques to maximize the compression rate while maintaining great accuracy. LLMs' Feed-Forward Network (FFN) components, which typically comprise a large proportion of parameters (around 2/3), ensure that our FFN optimizations would have a better chance of achieving effective compression. Moreover, our findings are beneficial to general LLMs and are not restricted to ReLU-based models. This work systematically investigates the tradeoff between enforcing activation sparsity and perplexity (accuracy) on state-of-the-art LLMs. Our empirical analysis demonstrates that we can obtain around 50% of main memory and computing reductions for critical FFN components with negligible accuracy degradation. This extra 50% sparsity does not naturally exist in the current LLMs, which require tuning LLMs' activation outputs by injecting zero-enforcing thresholds. To obtain the benefits of activation sparsity, we provide a guideline for the system architect for LLM prediction and prefetching. The success prediction allows the system to prefetch the necessary weights while omitting the inactive ones and their successors, therefore lowering cache and memory pollution and reducing LLM execution time on resource-constrained edge devices.

cs.LG↗

arXiv2024

The Robustness of Spiking Neural Networks in Communication and its Application towards Network Efficiency in Federated Learning

Spiking Neural Networks (SNNs) have recently gained significant interest in on-chip learning in embedded devices and emerged as an energy-efficient alternative to conventional Artificial Neural Networks (ANNs). However, to extend SNNs to a Federated Learning (FL) setting involving collaborative model training, the communication between the local devices and the remote server remains the bottleneck, which is often restricted and costly. In this paper, we first explore the inherent robustness of SNNs under noisy communication in FL. Building upon this foundation, we propose a novel Federated Learning with Top-K Sparsification (FLTS) algorithm to reduce the bandwidth usage for FL training. We discover that the proposed scheme with SNNs allows more bandwidth savings compared to ANNs without impacting the model's accuracy. Additionally, the number of parameters to be communicated can be reduced to as low as 6 percent of the size of the original model. We further improve the communication efficiency by enabling dynamic parameter compression during model training. Extensive experiment results demonstrate that our proposed algorithms significantly outperform the baselines in terms of communication cost and model accuracy and are promising for practical network-efficient FL with SNNs.

cs.LG↗

arXiv2016

A Brief Survey of Non-Residue Based Computational Error Correction

The idea of computational error correction has been around for over half a century. The motivation has largely been to mitigate unreliable devices, manufacturing defects or harsh environments, primarily as a mandatory measure to preserve reliability, or more recently, as a means to lower energy by allowing soft errors to occasionally creep. While residue codes have shown great promise for this purpose, there have been several orthogonal non-residue based techniques. In this article, we provide a high level outline of some of these non-residual approaches.

cs.OH↗