A meta-subgraph ( M ), defined as an edge-unfolded subgraph consisting of a set of connected edge types in the schema of a heterogeneous information network (HIN), generalises the classical meta-path. Although prior studies have leveraged instances of special shapes defined on the schema for tasks such as cohesive subgraph discovery, similarity measurement and recommendation, no existing work has formally defined the M -instance search problem and developed efficient algorithms dedicated to it. In this paper, we first generalise M queries and then systematically explore efficient algorithms dedicated to M instance search. For a given query M , the vertex/edge correspondences are fixed by the schema. Thus, we do not search for embeddings; instead, we directly retrieve the corresponding instance subgraphs. This avoids generating embedding permutations, although the decision version remains NP-complete. We propose two baselines adapted from backtracking and degeneracy-ordering strategies, which scale only to small HINs. We then derive a new upper bound on the number of M instances, | Ans | * edge , a specialisation of the AGM bound, which exploits edge-type adjacency in M and typed edge lists. Guided by the idea of the bound, we enhance backtracking with effective pruning, achieving complexity O * (| Ans | * edge ). Extending and generalising this analysis to subgraphs of M adjacency and instances of subgraphs of M yields | Ans | * bag . By fully considering the trade-off over: 1) the cost of deriving optimum subqueries, 2) the accuracy and cost of estimating the number of instances of subqueries, and 3) the cost of materialising instances of subqueries, we propose to utilise star shape subgraphs of M to search M instances, which runs in O * (| Ans | * star ) and is practically much faster due to cross-star based pruning. Experiments on large real-world HINs demonstrate that our best algorithm runs up to two orders of magnitude faster than the baselines and direct AGM-based approaches.
Dynamic graphs, pivotal in applications ranging from social networks to biological systems, pose significant challenges in storage and update efficiency due to their continuous evolution through insertions and deletions. Traditional graph compression methods, primarily optimized for static graphs, often struggle in dynamic environments, leading to costly decompression-recompression cycles during updates. To address this limitation, we propose a novel theoretical framework that enables efficient direct updates on rule-based compressed graphs. Additionally, we introduce a dynamic graph processing framework that balances space efficiency, update responsiveness, and query performance. Our framework treats updates as lightweight, localized modifications, sustains compression via background cleanup, and preserves query performance by ensuring structural integrity. Our method achieves significant memory savings, reducing usage by an average of 50.1% over state-of-the-art dynamic compressed graph systems, while demonstrating highly competitive update throughput and query performance. These advancements establish a scalable, high-performance solution for dynamic graph management, effectively bridging the gap between space efficiency, updates responsiveness and query performance.
Ever since the Dennard scaling broke down in the early 2000s and the frequency of the CPUs stalled, vendors have started to increase the core count in each CPU chip at the expense of introducing heterogeneity, thus ushering the era of NUMA and Chiplet processors. Since then, the heterogeneity in the design space of hardware has only increased to the point that DBMS performance may vary significantly up to an order of magnitude in modern servers. An important factor that affects performance includes the location of the logical cores where the DBMS queries execute, and the location where the data resides. This paper introduces P-MOSS, a learned spatial scheduling framework that schedules query execution to specific logical cores, and co-locates data on the corresponding NUMA node. For cross-hardware and workload adaptability, P-MOSS leverages core principles from Large Language Models, such as Next Token prediction, Generative Pre-training, and Fine-tuning. In the spirit of hardware-software synergy, P-MOSS guides its scheduling decision solely based on the low-level hardware statistics collected from the hardware Performance Monitoring Unit with the aid of a Decision Transformer. Experimental evaluation is performed in the context of the B+-Tree index. Performance results demonstrate that P-MOSS offers an improvement of up to 6 & times; over traditional schedules in terms of query throughput.
Federated learning (FL) has emerged as a popular paradigm for distributed machine learning over decentralized data. A typical FL training task involves a fleet of client devices with private data and a centralized server for aggregating the global model. Data generated by FL clients, e.g., smart phones, vehicles, and cameras, is prone to noise. While the impact of data noise on centralized learning (CL) is well understood, to our best knowledge there is a lack of a systematic study from this point of view for FL. In this paper, we fill this gap by presenting an empirical investigation to provide a deeper understanding regarding the impact of data noise on FL. Our study is enabled by DataNoiseGenerator, an open-source and extensible toolkit that we developed for the injection of controlled data noise across five diverse data modalities: image, video, audio, text, and tabular data. We then carry out extensive experiments based on the noisy data generated by DataNoiseGenerator, and our experimental evaluation results reveal that FL is significantly more vulnerable to data noise compared to CL, in terms of the quality of the trained ML models. This gap between FL and CL widens as the intensity of data noise and the proportion of noisy FL clients increase. We further present a detailed analysis to diagnose the root cause of this increased sensitivity of FL to data noise. Our analysis finds that the aggregation performed by the FL server can amplify divergent updates from FL clients trained on noisy data, thereby hindering global model convergence. We conclude that data quality issues are a fundamental challenge for deploying robust FL systems and demand novel decentralized data cleaning mechanisms.
We introduce De Finetti Logic, a novel formalization of the data model of Gamma Probabilistic Databases. De Finetti Logic is a simple, database-centric probabilistic programming framework in which models are specified through relational constraints applied to a probabilistic database. This framework is grounded in the concept of Polya urns and offers an intuitive formalization of exchangeability, the fundamental structural assumption underlying Gamma Probabilistic Databases. We leverage this new formalism to develop a novel inference mechanism based on variational methods. Starting from a DFL theory, our inference method automatically identifies an appropriate family of variational distributions to approximate the target posterior, and synthesizes an optimization algorithm to minimize the Kullback-Leibler divergence between the true posterior and its variational surrogate. To maximize performance, we employ knowledge compilation techniques to limit the number of variational parameters, while preserving the expressive power of the surrogate. Our method is implemented as an extension to the StarfishDB system. Experimental results demonstrate that (i) our approach provides a practical, scalable, and fast-converging alternative to general-purpose inference via collapsed Gibbs sampling, and (ii) remains competitive with specialized algorithms that leverage model-specific optimizations.
