arXiv subjects

Danni Peng

Publications and source records attributed to Danni Peng.

10 recordsLinked to original sources

arXiv2026

Enabling High-Bandwidth Flash for Generative Recommendation Serving with Write-Aware KV Cache Policy

Generative recommendation (GR) systems increasingly leverage user-level KV cache reuse to avoid recomputing long user histories. However, the growing KV cache capacity and bandwidth requirements introduce new challenges for memory system. High-Bandwidth Flash (HBF) provides a promising solution by offering substantially higher capacity than HBM while approaching HBM-class read bandwidth, enabling larger scale KV cache retention and improved serving throughput. Yet conventional Least-Recently-Used (LRU) KV cache management tightly couples KV cache writes with cache misses, generating excessive write traffic that rapidly exhausts flash endurance. In this work, we evaluate a write-aware KV cache policy based on admission-controlled LRU-K for HBF-based GR serving. By filtering low-reuse users before cache admission, LRU-K decouples KV cache writes from misses and significantly reduces unnecessary writes. We develop an analytical model to characterize GR serving performance, KV cache write traffic, and HBF lifetime, and evaluate performance across diverse memory systems and GR workloads. Our results show that HBF-based systems achieve 3.8 to 4.7 times higher throughput than HBM-only systems. Moreover, LRU-K extends HBF lifetime from about one year under conventional LRU to over six years with a moderate K=10, while maintaining comparable or even slightly improved throughput. These results highlight the importance of write aware KV cache policy for sustainable HBF-based GR serving.

cs.AR↗

arXiv2026

Ramsey properties for tilings in random graphs

Let $mH$ be the graph formed by $m$ vertex-disjoint copies of a graph $H$. Let $G \to (H)_r$ denote that, in any $r$-colouring of the edges of $G$, there exists a monochromatic copy of $H$. In 1975, Burr, Erd\H{o}s, and Spencer showed that if $H$ is a graph on $k$ vertices whose independence number is $\alpha$, then $K_n \to (mH)_2$, where $m\sim n/(2k-\alpha)$, and that the $1/(2k-\alpha)$ factor is best possible. In the 1990s, R\"{o}dl and Ruci\'{n}ski proved that, for all but a few graphs~$H$, the threshold for the property $\mathbb{G}(n,p) \to (H)_r$ is $n^{-1/m_2(H)}$. In this paper, generalizing the result of Burr, Erd\H{o}s, and Spencer, we prove that $n^{-1/\max\{m_2(H),1\}}$ is the threshold for the property $\mathbb{G}(n,p) \to (mH)_2$, where $m\sim n/(2k-\alpha)$. This threshold matches the one found by R\"{o}dl and Ruci\'nski for most graphs $H$, extending their result in the case $r=2$.

math.CO↗

arXiv2025

Separating path systems for cubic graphs and for complete bipartite graphs

A strongly separating path system in a graph $G$ is a collection $\mathcal{P}$ of paths in $G$ such that, for every two edges $e$ and $f$ of $G$, there is a paths in $\mathcal{P}$ with $e$ and not $f$, and vice-versa. The minimum number of such a system is the so called strong separation number of $G$. We prove that the strong separation number of every $2$-degenerate graph on $n$ vertices is at most $n$. Using this, we also provide upper bounds for the strong separation number of subcubic graphs, planar graphs, and planar bipartite graphs. On the other hand, we prove that the strong separation number a complete bipartite graph $K_{a,b}$ is at least $b$ if $a<b/2$ and at least $(\sqrt{6(b/2)+4}-2)a$ if $b/2\leq a\leq b$, and we provide a construction that attains the former bound.

math.CO↗

arXiv2025

Improving Learning of New Diseases through Knowledge-Enhanced Initialization for Federated Adapter Tuning

In healthcare, federated learning (FL) is a widely adopted framework that enables privacy-preserving collaboration among medical institutions. With large foundation models (FMs) demonstrating impressive capabilities, using FMs in FL through cost-efficient adapter tuning has become a popular approach. Given the rapidly evolving healthcare environment, it is crucial for individual clients to quickly adapt to new tasks or diseases by tuning adapters while drawing upon past experiences. In this work, we introduce Federated Knowledge-Enhanced Initialization (FedKEI), a novel framework that leverages cross-client and cross-task transfer from past knowledge to generate informed initializations for learning new tasks with adapters. FedKEI begins with a global clustering process at the server to generalize knowledge across tasks, followed by the optimization of aggregation weights across clusters (inter-cluster weights) and within each cluster (intra-cluster weights) to personalize knowledge transfer for each new task. To facilitate more effective learning of the inter- and intra-cluster weights, we adopt a bi-level optimization scheme that collaboratively learns the global intra-cluster weights across clients and optimizes the local inter-cluster weights toward each client's task objective. Extensive experiments on three benchmark datasets of different modalities, including dermatology, chest X-rays, and retinal OCT, demonstrate FedKEI's advantage in adapting to new diseases compared to state-of-the-art methods.

cs.LG↗

arXiv2025

Look Back for More: Harnessing Historical Sequential Updates for Personalized Federated Adapter Tuning

Personalized federated learning (PFL) studies effective model personalization to address the data heterogeneity issue among clients in traditional federated learning (FL). Existing PFL approaches mainly generate personalized models by relying solely on the clients' latest updated models while ignoring their previous updates, which may result in suboptimal personalized model learning. To bridge this gap, we propose a novel framework termed pFedSeq, designed for personalizing adapters to fine-tune a foundation model in FL. In pFedSeq, the server maintains and trains a sequential learner, which processes a sequence of past adapter updates from clients and generates calibrations for personalized adapters. To effectively capture the cross-client and cross-step relations hidden in previous updates and generate high-performing personalized adapters, pFedSeq adopts the powerful selective state space model (SSM) as the architecture of sequential learner. Through extensive experiments on four public benchmark datasets, we demonstrate the superiority of pFedSeq over state-of-the-art PFL methods.

cs.LG↗

arXiv2024

Near rainbow Hamilton cycles in dense graphs

Finding near-rainbow Hamilton cycles in properly edge-coloured graphs was first studied by Andersen, who proved in 1989 that every proper edge colouring of the complete graph on $n$ vertices contains a Hamilton cycle with at least $n-\sqrt{2n}$ distinct colours. This result was improved to $n-O(\log^2 n)$ by Balogh and Molla in 2019. In this paper, we consider Anderson's problem for general graphs with a given minimum degree. We prove every globally $n/8$-bounded (i.e. every colour is assigned to at most $n/8$ edges) properly edge-coloured graph $G$ with $\delta(G) \geq (1/2+\varepsilon)n$ contains a Hamilton cycle with $n-o(n)$ distinct colours. Moreover, we show that the constant $1/8$ is best possible.

math.CO↗

arXiv2023

More on Rainbow Cliques in Edge-Colored Graphs

In an edge-colored graph $G$, a rainbow clique $K_k$ is a $k$-complete subgraph in which all the edges have distinct colors. Let $e(G)$ and $c(G)$ be the number of edges and colors in $G$, respectively. In this paper, we show that for any $\varepsilon>0$, if $e(G)+c(G) \geq (1+\frac{k-3}{k-2}+2\varepsilon) {n\choose 2}$ and $k\geq 3$, then for sufficiently large $n$, the number of rainbow cliques $K_k$ in $G$ is $\Omega(n^k)$. We also characterize the extremal graphs $G$ without a rainbow clique $K_k$, for $k=4,5$, when $e(G)+c(G)$ is maximum. Our results not only address existing questions but also complete the findings of Ehard and Mohr (Ehard and Mohr, Rainbow triangles and cliques in edge-colored graphs. {\it European Journal of Combinatorics, 84:103037,2020}).

math.CO↗

arXiv2022

Learning Gradient-based Mixup with Extrapolation toward Flatter Minima for Domain Generalization

To address distribution shifts between training and test data, domain generalization (DG) leverages multiple source domains to learn a model that generalizes well to unseen domains. However, existing DG methods often overfit to the source domains, partly due to the limited coverage of the expected region in feature space. Motivated by this, we propose performing mixup with data interpolation and extrapolation to cover potentially unseen regions. To prevent the detrimental effects of unconstrained extrapolation, we carefully design a policy to generate the instance weights, named Flatness-aware Gradient-based Mixup (FGMix). The policy relies on gradient-based compatibilities to assign greater weights to instances that carry more invariant information and learn the mixup policy towards flatter minima for better generalization. On the DomainBed benchmark, we validate the efficacy of various designs of FGMix and demonstrate its superiority over other DG algorithms.

cs.LG↗

arXiv2021

Tur\'an Number of Subdivisions of Multipartite Graphs

In this paper, we investigate the Tur\'an exponent for $1$-subdivisions of graphs that are neither bipartite nor complete. Specifically, we establish an upper bound on the Tur\'an number of the 1-subdivision of $K_{s,t}^+$, where $K_{s,t}^+$ is obtained by adding a single edge within the part of size $s$ of the complete bipartite graph $K_{s,t}$, with $4\leq s \leq t$. In addition, we derive an upper bound for the extremal number of a family of graphs formed by (possibly degenerate) 1-subdivisions of certain tripartite graphs.

math.CO↗

arXiv2021

Learning an Adaptive Meta Model-Generator for Incrementally Updating Recommender Systems

Recommender Systems (RSs) in real-world applications often deal with billions of user interactions daily. To capture the most recent trends effectively, it is common to update the model incrementally using only the newly arrived data. However, this may impede the model's ability to retain long-term information due to the potential overfitting and forgetting issues. To address this problem, we propose a novel Adaptive Sequential Model Generation (ASMG) framework, which generates a better serving model from a sequence of historical models via a meta generator. For the design of the meta generator, we propose to employ Gated Recurrent Units (GRUs) to leverage its ability to capture the long-term dependencies. We further introduce some novel strategies to apply together with the GRU meta generator, which not only improve its computational efficiency but also enable more accurate sequential modeling. By instantiating the model-agnostic framework on a general deep learning-based RS model, we demonstrate that our method achieves state-of-the-art performance on three public datasets and one industrial dataset.

cs.IR↗