Federated learning (FL) allows model training from local data collected by edge/mobile devices while preserving privacy, which has wide applicability to image and vision applications. A challenge is that client in FL usually have much more limited computation communication resources compared servers a center. To overcome this challenge, we propose PruneFL -a novel approach with adaptive distributed parameter pruning, adapts the size during reduce both overhead minimize overall time,...

Federated learning (FL) allows model training from local data collected by edge/mobile devices while preserving privacy, which has wide applicability to image and vision applications. A challenge is that client in FL usually have much more limited computation communication resources compared servers a datacenter. To overcome this challenge, we propose PruneFL -- novel approach with adaptive distributed parameter pruning, adapts the size during reduce both overhead minimize overall time,...

In many real applications, especially those involving data objects with complicated semantics, it is generally desirable to discover the relation between patterns in input space and labels corresponding different semantics output space. This task becomes feasible MIML (Multi-Instance Multi-Label learning), a recently developed learning framework, where each object represented by multiple instances allowed be associated simultaneously. this paper, we propose KISAR, an algorithm that able what...

In the domain of Natural Language Processing (NLP), Large Models (LLMs) have demonstrated promise in text-generation tasks. However, their educational applications, particularly for domain-specific queries, remain underexplored. This study investigates LLMs' capabilities scenarios, focusing on concept graph recovery and question-answering (QA). We assess zero-shot performance creating graphs introduce TutorQA, a new expert-verified NLP-focused benchmark scientific reasoning QA. TutorQA...

Purpose The Shanghai Preconception Cohort (SPCC) was initially established to investigate the associations of parental periconceptional nutritional factors with congenital heart disease (CHD) but has further analysed child growth and development paediatric diseases. Participants Preparing-for-pregnancy couples who presented at preconception examination clinics early-pregnancy women before 14 gestational weeks were enrolled comprise baseline study population. General characteristics, routine...

Dynamic resource allocation to satisfy varying, concurrent and unpredictable demands from multiple applications is a key need in cloud systems. A fundamental challenge the find right balance between over-allocation, which satisfies each application's varying needs without requiring frequent changes, system efficiency requires that exactly matches application needs. However, allocating resources close current will result changes. This can be detrimental since there may fixed costs (state...

Efficient Transformers have been developed for long sequence modeling, due to their subquadratic memory and time complexity. Sparse Transformer is a popular approach improving the efficiency of by restricting self-attention locations specified predefined sparse patterns. However, leveraging sparsity may sacrifice expressiveness compared full-attention, when important token correlations are multiple hops away. To combine advantages both transformer full-attention Transformer, we propose...

Knowledge Graphs (KGs) are crucial in the field of artificial intelligence and widely used downstream tasks, such as question-answering (QA). The construction KGs typically requires significant effort from domain experts. Large Language Models (LLMs) have recently been for Graph Construction (KGC). However, most existing approaches focus on a local perspective, extracting knowledge triplets individual sentences or documents, missing fusion process to combine global KG. This work introduces...

In this paper, we introduce theTheory of Bottleneck Ordering, a mathematical framework that reveals the bottleneck structure data networks. This theoretical provides insights into inherent topological properties network in at least three areas: (1) It identifies regions influence each bottleneck; (2) it order which bottlenecks (and flows traversing them) converge to their steady state transmission rates distributed congestion control algorithms; and (3) key design optimized traffic...

This paper provides a mathematical model of data center performance based on the recently introduced Quantitative Theory Bottleneck Structures (QTBS). Using model, we prove that if traffic pattern is \textit{interference-free}, there exists unique optimal design both minimizes maximum flow completion time and yields maximal system-wide throughput. We show interference-free patterns correspond to important set display locality properties use these theoretical insights study three widely used...

In this paper, we introduce the Theory of Bottleneck Ordering, a mathematical framework that reveals bottleneck structure data networks. This theoretical provides insights into inherent topological properties network in at least three areas: (1) It identifies regions influence each bottleneck; (2) it order which bottlenecks (and flows traversing them) converge to their steady state transmission rates distributed congestion control algorithms; and (3) key design optimized traffic engineering...

Educational materials such as survey articles in specialized fields like computer science traditionally require tremendous expert inputs and are therefore expensive to create update. Recently, Large Language Models (LLMs) have achieved significant success across various general tasks. However, their effectiveness limitations the education domain yet be fully explored. In this work, we examine proficiency of LLMs generating succinct specific niche field NLP science, focusing on a curated list...

Large language models (LLMs) have been a disruptive innovation in recent years, and they play crucial role our daily lives due to their ability understand generate human-like text. Their capabilities include natural understanding, information retrieval search, translation, chatbots, virtual assistance, many more. However, it is well known that LLMs are massive terms of the number parameters. Additionally, self-attention mechanism underlying architecture LLMs, Transformers, has quadratic...

Knowledge graphs (KGs) are crucial in the field of artificial intelligence and widely applied downstream tasks, such as enhancing Question Answering (QA) systems. The construction KGs typically requires significant effort from domain experts. Recently, Large Language Models (LLMs) have been used for knowledge graph (KGC), however, most existing approaches focus on a local perspective, extracting triplets individual sentences or documents. In this work, we introduce Graphusion, zero-shot KGC...

360-degree panoramic videos have gained considerable attention in recent years due to the rapid development of head-mounted displays (HMDs) and cameras. One major problem streaming is that are much larger size compared traditional ones. Moreover, user devices often a wireless environment, with limited battery, computation power, bandwidth. To reduce resource consumption, researchers proposed ways predict users' viewports so only part entire video needs be transmitted from server. However,...

Congestion control algorithms for data networks have been the subject of intense research last three decades. While most work has focused around characterization a flow's bottleneck link, understanding interactions amongst links and ripple effects that perturbations in link can cause on rest network remained much less understood. The Theory Bottleneck Ordering is recently developed mathematical framework reveals structure provides model to understand such effects. In this paper we present...

In this paper,we introduce the Theory of Bottleneck Ordering, a mathematical framework that reveals bottleneck structure data networks. This theoretical provides insights into inherent topological properties network in at least three areas: (1) It identifies regions influence each bottleneck; (2) it order which bottlenecks (and flows traversing them) converge to their steady state transmission rates distributed congestion control algorithms; and (3) key design optimized traffic engineering...

The conventional view of the congestion control problem in data networks is based on principle that a flow's performance uniquely determined by state its bottleneck link, regardless topological properties network. However, recent work has shown behavior congestion-controlled better explained models account for interactions between links. These are captured latent \textit{bottleneck structure}, model describing complex ripple effects changes one part network exert other parts. In this paper,...

This paper uses models such as factor analysis for multidimensional information of survey subjects. Through literature research and the actual situation, we identify factors that public tends to ignore about energy saving emission reduction, construct an econometric model get specific mathematical expressions influencing factors, find out with a higher degree importance. And based on results, propose targeted suggestions.

Accurate Photovoltaic power (PV) forecasting is the basis and key to grid dispatch management. With machine learning algorithms latest swarm intelligence being proposed, a reasonable combination of two will produce good prediction results. This paper addresses problem optimal selection hyperparameters for XGBoost algorithm in PV problem. establishes an long-term model based on optimization ISMA algorithm, firstly, dataset pre-processed training set test are divided, then data trained, with...

This paper presents a sufficient condition for stochastic gradients not to slow down the convergence of Nesterov's accelerated gradient method. The new has strong-growth by Schmidt \& Roux as special case, and it also allows us (i) model problems with constraints (ii) design types oracles (e.g., finite-sum such SAGA). Our results are obtained revisiting algorithm useful designing without changing underlying first-order