Yu Cheng

Congratulations to Qianlin Liang on successfully defending his Ph.D thesis titled "Resource Management for Edge AI." Qianlin's thesis looks at model-driven resource management for edge AI workloads. #edgeAI #cloudcomputing

93

7 Comments

Computing high-degree polynomial gradients in memory https://lnkd.in/gUWXbXDV

1

[Image linked] Buckle up, tech enthusiasts! 🚀 Our understanding of neural networks is getting a whole new framework, thanks to Jiaming Luo. This graduate student's intriguing discovery sets the stage for Liquid #AI - a game-changer that's turning static neural networks dynamic. Let's applaud this innovation that could revolutionize data interpretation.👏🎓 But, predictably, this shift brings unique challenges regarding transparency and accountability. Want to dive into the depths of this compelling breakthrough? Follow the link to satiate your curiosity: 👉 https://lnkd.in/en9A-UiZ #ArtificialIntelligence #NeuralNetworks #TechnologyNews

2

An interesting article published by International researchers (National Yang Ming Chiao Tung University, Chung Hua University, National Ilan University) on #ScienceProgress point out a new multimodal fusion model that integrates both textual and visual information to challenging the #fakenews phenomenon. Unlike traditional single-modal detection models, which often analyze text or images separately, the proposed approach utilizes early fusion, joint fusion, and late fusion techniques to enhance detection accuracy. The study achieves notable accuracy rates (85% to 90%) using datasets like #Gossipcop and #Fakeddit, showing improved fake news detection capabilities through multimodal fusion. 🌐 Reference: https://lnkd.in/dxfwpcsE 👥 Szu-Yin Lin, Yen-Chiu Chen, Kuo-Ming Chao, ... 🗣 “Our results demonstrate that the joint fusion method not only improves accuracy but provides greater flexibility for real-world application.” "Text–image multimodal fusion model for enhanced fake news detection" is structured into #sections covering the introduction, methodology, experiments, results, and conclusion. It provides a detailed explanation of the research framework, data preprocessing, feature extraction, fusion methods, classification models, and evaluation metrics and as results it demonstrates that multimodal fusion significantly improves fake news detection accuracy and F1-score: 1️⃣ Introduction: Background on fake news proliferation, relevance of multimodal detection; 2️⃣ Related work: Review of single-modal fake news detection and early multimodal studies; 3️⃣ Methodology: Detailed explanation of data preprocessing, feature extraction, fusion strategies, and classification; 4️⃣ Experiments and Results: Empirical performance of models on Gossipcop and Fakeddit datasets; 5️⃣ Discussion: Comparative analysis of fusion strategies, limitations, and recommendations; 6️⃣ Conclusion and future directions: Summary of findings and proposed future research. As #result of the study, early fusion increased accuracy and F1-score by up to 15% and 17%, respectively, while late fusion further improved accuracy by 3% and F1-score by 4%. The proposed model achieved F1-scores of 90% and 88% for the Gossipcop and Fakeddit datasets, respectively. 💡 As further research fields, authors highlight: ➡️ Domain-independent datasets to evaluate generalization; ➡️ Enhanced fusion strategies to address limitations in context-based detection; ➡️ Fine-tuning multimodal techniques to optimize for high variability in social media content. #ArtificialIntelligence #Innovation #Research #AI #News #Fake

18

DRESS Kit V2 is finally here! This updated version retains familiar method signatures, making it easier for researchers to transition while unlocking new capabilities. Read more about it in Waihong Chung's article. #MachineLearning #Programming

3

💥💥💥 Can Graph Learning Improve Planning in LLM-based Agents? Xixi Wu, Yifei Shen, Caihua Shan, Kaitao Song, Siwei Wang, Bohang Zhang, Jiarui Feng, Hong Cheng, Wei Chen, Yun Xiong, Dongsheng Li Abstract Task planning in language agents is emerging as an important research topic alongside the development of large language models (LLMs). It aims to break down complex user requests in natural language into solvable sub-tasks, thereby fulfilling the original requests. In this context, the sub-tasks can be naturally viewed as a graph, where the nodes represent the sub-tasks, and the edges denote the dependencies among them. Consequently, task planning is a decision-making problem that involves selecting a connected path or subgraph within the corresponding graph and invoking it. In this paper, we explore graph learning-based methods for task planning, a direction that is orthogonal to the prevalent focus on prompt design. Our interest in graph learning stems from a theoretical discovery: the biases of attention and auto-regressive loss impede LLMs' ability to effectively navigate decision-making on graphs, which is adeptly addressed by graph neural networks (GNNs). This theoretical insight led us to integrate GNNs with LLMs to enhance overall performance. Extensive experiments demonstrate that GNN-based methods surpass existing solutions even without training, and minimal training can further enhance their performance. The performance gain increases with a larger task graph size. 👉 https://lnkd.in/dMC5t6Wg #machinelearning

1

Witness a landmark moment in AI: Dr Yang Liu (HKBU Department of Computer Science) originated a pioneering polarised message-passing (PMP) framework, featured in the prestigious journal of 𝘈𝘳𝘵𝘪𝘧𝘪𝘤𝘪𝘢𝘭 𝘐𝘯𝘵𝘦𝘭𝘭𝘪𝘨𝘦𝘯𝘤𝘦. Dr Liu debuted three innovative PMP-augmented graph neural networks (GNNs), surpassing leading methodologies in diverse real-world applications, and forging new industry benchmarks with unparalleled precision and efficiency. 𝐑𝐞𝐚𝐝 𝐭𝐡𝐞 𝐏𝐚𝐩𝐞𝐫: https://lnkd.in/gTKpBqpc      🗝 Cluster: Data Analytics and Artificial Intelligence #Cluster_Data_Analytics_and_Artificial_Intelligence #HKBU #ResearchOffice #HKBUResearch #HKBURO #HKBUCOMP #ComputerScience #PolarizedMessagePassing #GraphNeuralNetwork #GNN #AIJ #ArtificialIntelligence #AI #Elsevier Elsevier

18

📌 How would you like to know the AI expert, Dr. Andrew Ng's insights on the next wave of AI trends for Taiwan's industries? For those who missed the live forum, you're in luck! Here’s your chance to catch up! The highly influential figure in AI (TIME100 AI), Silicon Valley AI leader, startup pioneer, and Managing General Partner of AI Fund, Dr. Andrew Ng, has brought his keynote speech titled AI, Agents, and Applications online!! 🥳 Don’t miss this chance to explore the latest AI trends — tune in and share your insights in the comments! 💬 🔻 Watch the Keynote🔻 https://reurl.cc/6dpQb6 #GenAI #AIAgents #ai #AIFund #LandingAI #AndrewNg #AgenticAI

25

💥💥💥 Towards Knowledge Checking in Retrieval-augmented Generation: A Representation Perspective Shenglai Zeng, Jiankun Zhang, Bingheng Li, Yuping Lin, Tianqi Zheng, Dante Everaert, Hanqing Lu, Hui Liu, Hui Liu, Yue Xing, Monica Xiao Cheng, Jiliang Tang Abstract Retrieval-Augmented Generation (RAG) systems have shown promise in enhancing the performance of Large Language Models (LLMs). However, these systems face challenges in effectively integrating external knowledge with the LLM's internal knowledge, often leading to issues with misleading or unhelpful information. This work aims to provide a systematic study on knowledge checking in RAG systems. We conduct a comprehensive analysis of LLM representation behaviors and demonstrate the significance of using representations in knowledge checking. Motivated by the findings, we further develop representation-based classifiers for knowledge filtering. We show substantial improvements in RAG performance, even when dealing with noisy knowledge databases. Our study provides new insights into leveraging LLM representations for enhancing the reliability and effectiveness of RAG systems. 👉 https://lnkd.in/dwpbPFfW #machinelearning

2

⛱️ Mixture of Experts Primer 🚀 The Mixture of Experts (MoE) model is revolutionizing large language models (LLMs) by enabling unprecedented scalability and efficiency. 🌟 🔗 Link to the primer on vinija.ai https://lnkd.in/gTS7eBzC . Written in collaboration with Aman Chadha MoE divides complex problems into smaller sub-tasks, each managed by specialized sub-models or "experts." This not only enhances performance but also optimizes computational resources by activating only the most relevant experts for each input. Key aspects: - Scalability: Linear increase in capacity without added computational complexity. - Efficiency: Activates only the necessary experts, reducing computation. - Flexibility: Adapts to various tasks with specialized experts. MoE is the driving force behind some of the most advanced AI models today: - Google’s Switch Transformer - OpenAI’s GPT-4 - Mistral AI’s Mixtral - Google’s Gemini 1.5 These advancements highlight MoE’s potential to create more powerful and efficient AI solutions. 🌍

103

4 Comments

Check out this #OA paper, “Research on Animation Character Expression Generation Based on Attention Conditioned Cyclegan”, by Chunling Kong (Zhengzhou University of Industrial Technology, China), featured in the Journal of Cases on Information Technology (JCIT). With advancements in computer vision and AI, deep learning is applied in animation and game design to enhance facial expression realism. This study proposes ACCGAN, an innovative method for automated animated character expression generation and transfer using deep learning. Access the full paper here > https://buff.ly/4eMhOsv #ZhengzhouUniversityofIndustrialTechnology #CharacterExpression #GenerativeAdversarialNetworks

1

Submit to Int. Journal of Activity and Behavior Computing (IJABC) Free to publish & open access - submit your quality works please, and inform others. Best regards IJABC @ijabc #ijabc Sozo Inoue Ahad VISION Lab (AI & ML: Vision2Future)

3

An efficient way to train more reliable #AI agents. Reinforcement learning (RL) is surprisingly brittle to contextual variations in tasks. A new method from LIDS PI Cathy Wu and LIDS grad students Jung-Hoon Cho, Vindula Jayawardana, and Sirui Li for solving contextual RL problems achieves 5-50x better sample efficiency on standard & traffic benchmarks. More at MIT News: https://bit.ly/40YSQC8 MIT Schwarzman College of Computing MIT Institute for Data, Systems, and Society (IDSS) MIT Civil and Environmental Engineering

19

We encourage you to delve into the recently published paper "Computation of Graph Fourier Transform Centrality Using Graph Filter". 📝 Authored by: Chien-Cheng Tseng; Su-Ling Lee In this paper, the computation of graph Fourier transform centrality (GFTC) of complex network using graph filter is presented. 🔗 Read more on IEEE Xplore: https://loom.ly/yKTKzAw #OpenAccess #OJCAS #OA #PaperHighlight

2