Knowledge Graph Embeddings (KGEs) support a wide range of downstream tasks over Knowledge Graphs (KGs). In practice, KGs evolve as new entities and facts are added, motivating Continual Knowledge Graph Embedding (CKGE) methods that update embeddings over time. Current CKGE approaches address catastrophic forgetting (i.e., the performance degradation on previously learned tasks) primarily by limiting changes to existing embeddings. However, we show that this view is incomplete. When new entities are introduced, their embeddings can interfere with previously learned ones, causing the model to predict them in place of previously correct answers. This phenomenon, which we call entity interference, has been largely overlooked and is not accounted for in current CKGE evaluation protocols. As a result, the assessment of catastrophic forgetting becomes misleading, and CKGE methods performance is systematically overestimated. To address this issue, we introduce a corrected CKGE evaluation protocol that accounts for entity interference. Through experiments on multiple benchmarks, we show that ignoring this effect can lead to performance overestimation of up to 25

The increasing demand for artificial intelligence (AI) workloads across diverse computing environments has driven the need for more efficient data management strategies. Traditional cloud-based architectures struggle to handle the sheer volume and velocity of AI-driven data, leading to inefficiencies in storage, computation, and data movement. This paper explores the integration of active storage systems within the computing continuum to optimize AI workload distribution.By embedding computation directly into storage architectures, active storage is able to reduce data transfer overhead, enhancing performance and improving resource utilization. Other existing frameworks and architectures offer mechanisms to distribute certain AI processes across distributed environments; however, they lack the flexibility and adaptability that the continuum requires, both regarding the heterogeneity of devices and the rapid-changing algorithms and models being used by domain experts and researchers.This article proposes a software architecture aimed at seamlessly distributing AI workloads across the computing continuum, and presents its implementation using mainstream Python libraries and dataClay, an active storage platform. The evaluation shows the benefits and trade-offs regarding memory consumption, storage requirements, training times, and execution efficiency across different devices. Experimental results demonstrate that the process of offloading workloads through active storage significantly improves memory efficiency and training speeds while maintaining accuracy. Our findings highlight the potential of active storage to revolutionize AI workload management, making distributed AI deployments more scalable and resource-efficient with a very low entry barrier for domain experts and application developers.

Abstract Artificial intelligence (AI) and data-driven applications are transforming industry and research across domains. Despite research advances, the management of the AI and data life cycle in organizations remains largely manual and ad hoc. At the same time, data spaces initiatives are emerging, providing new data-sharing opportunities but posing new challenges for organizations seeking to leverage all their potential. Tackling these challenges, this chapter describes the new framework proposed by the CyclOps project, which aims to enable interoperable and trustworthy automatic management, governance, and maintenance of the entire data life cycle for large-scale volumes of data generated in heterogeneous distributed sources. This supports the development and deployment of AI-based applications across both business and research contexts. CyclOps operationalizes the end-to-end data life cycle, placing at its core knowledge graphs, an established semantic formalism to represent data and metadata adhering to the FAIR Principles, while capturing relevant information to improve reproducibility, traceability, and explainability of the AI results. This core layer is complemented with tools for the automation of data management tasks; distributed data processing; AI tools, algorithms, and models; data space interoperability; and a human-centric interface. The chapter presents the key innovative propositions of CyclOps and its underlying technologies and illustrates them through selected use cases that highlight the general applicability of the approach.

In today's data-driven world, the ability to extract meaningful information from data is becoming essential for businesses, organizations and researchers. For this purpose, a wide range of tools and systems exists addressing data-related tasks, from data integration, preprocessing and modeling, to the interpretation and evaluation of the results. As data continues to grow in volume and complexity, there is an increasing need for advanced yet userfriendly tools, such as intelligent discovery assistants (IDAs) or automated machine learning (AutoML) systems, that facilitate the user's interaction with data. This enables non-expert users to effectively leverage powerful data analytics techniques. However, the use of these tools still requires non-trivial user input that cannot be anticipated from the analytical problem's data alone, but must be tailored to each individual user and their specific intents. To this end, this work explores the use of Knowledge Graphs (KG) as a foundational representation for capturing complex analytics workflows, as well as information about the users, their intents and their feedback, in order to facilitate user interaction with IDAs or AutoML tools. This is achieved through established techniques from recommender systems, in particular a link prediction approach based on KG embeddings and graph neural networks. Experimental results show that the proposed method effectively captures the graph structure and produces meaningful suggestions for users. To demonstrate the feasibility of the approach, a working prototype is presented.

Knowledge Graphs (KGs) represent knowledge as structured graphs of entities and their relationships. Knowledge Graph Embeddings (KGEs) enable learning and inference over KGs by representing their entities and relations as continuous vectors. Many KGs are frequently updated, requiring their KGEs to adapt to these changes. Continual learning methods for KGEs address this by incorporating new embeddings while updating existing ones to account for the new facts without retraining from scratch. A necessary step in these methods is the initialization of new embeddings, which has an important impact both on the accuracy of the resulting embeddings, and in the time required for their training. Current continual learning methods either use a random initialization as in non-continual scenarios, ignoring relevant information in existing embeddings, or rely on a model-dependent initialization, which is only compatible with specific KGE training models. We propose a novel informed embedding initialization strategy that can be seamlessly integrated into existing continual learning methods for KGEs. Our approach leverages the KG schema along with previously learned embeddings to generate semantically informed initial representations at every continual learning step, enhancing knowledge acquisition while reducing catastrophic forgetting in the final embeddings obtained. Experimental results show that our strategy improves the predictive performance of the resulting KGEs and enhances knowledge retention. Moreover, it accelerates the learning process, reducing the number of epochs, and thus the time required to incrementally learn new embeddings. Finally, our evaluation demonstrates that these benefits are consistent regardless of the KGE learning model used.

Denial Constraints (DCs) form a highly expressive integrity rule language that subsumes many used formalisms such as keys and functional dependencies, making them widely adopted in applications that require the manipulation of rich sets of data constraints. This expressiveness has motivated the development of numerous algorithms for automatically discovering DCs from data, with particular emphasis on the discovery of approximate DCs to improve robustness to erroneous data. However, existing DC discovery algorithms exhibit computational costs that are quadratic in the number of tuples and exponential in the number of attributes, and most cannot accommodate changes in the data. Moreover, they often produce thousands of uninformative DCs. These limitations make current DC discovery algorithms difficult to use effectively on very large and dynamic databases. In this paper, we present LIMA, an approximate DC discovery algorithm that efficiently discovers DCs on very large and dynamic databases. LIMA uses statistical methods to infer properties of DCs from reduced samples, and introduces a novel discovery framework that exploits a more restrictive definition of DC validity to substantially reduce the cost of searching for valid DCs. We experimentally demonstrate that LIMA achieves significantly better scalability than current algorithms with respect to both rows and attributes, while also discovering higher-quality sets of DCs with precisions several orders of magnitude higher than the state of the art, both in static and in dynamic datasets.

We study the problem of efficiently computing rankings of joinable attributes in data lakes. Traditional set-overlap measures produce numerous false positives in this scenario, while modern, more accurate Table Representation Learning (TRL) techniques incur prohibitive computational costs. In contrast to the state-of-the-art, we adopt a novel notion of join quality tailored to data lakes relying on a metric that combines multiset Jaccard and cardinality proportion. The proposed metric merges the best of both worlds by leveraging syntactic measures while achieving accuracy scores comparable to those of TRL approaches. Generating rankings of joinable pairs is highly scalable at both preparation and query time, since we train a general-purpose predictive model. Predictions are based on data profiles, succinct and efficiently computed representations of dataset characteristics. Our experiments show that our system, Freyja, matches and improves upon, the results obtained by the state-of-the-art while reducing execution costs by orders of magnitude.

Despite advancements in MLOps and AutoML, ML development still remains challenging for data scientists. First, there is poor support for and limited control over optimizing and evolving ML models. Second, there is lack of efficient mechanisms for continuous evolution of ML models which would leverage the knowledge gained in previous optimizations of the same or different models. We propose an experiment-driven MLOps approach which tackles these problems. Our approach relies on the concept of an experiment, which embodies a fully controllable optimization process. It introduces full traceability and repeatability to the optimization process, allows humans to be in full control of it, and enables continuous improvement of the ML system. Importantly, it also establishes knowledge, which is carried over and built across a series of experiments and allows for improving the efficiency of experimentation over time. We demonstrate our approach through its realization and application in the ExtremeXP1 project (Horizon Europe).

Data spaces have recently emerged as an innovative paradigm for cross-organizational data sharing. These decentralized environments require sophisticated data governance protocols to ensure compliance with data standards, roles and policies. While current policy-based solutions address enforcement of data access control and usage rights, they lack mechanisms for automated data validation -essential for ensuring data quality for collaborative analytics. To address this gap, we present a knowledge graph-based framework to automate data validation inline with data policies. This framework relies on the concept of policy checkers, which represent high-level and technology-agnostic data validation plans that can be dynamically translated into technology-specific user defined functions (UDFs) for compliance checking. Importantly, the usage of knowledge graphs to describe the policy checkers enhances the transparency and traceability of data validation processes, while the two-stage process (technology-agnostic policy checkers and technology-specific UDFs) accommodate data validation on multimodal data. We accompany the description of our approach with a proof of concept that demonstrates the feasibility of this solution in real data spaces.

Many Knowledege Graphs (KGs) are frequently updated, forcing their Knowledge Graph Embeddings (KGEs) to adapt to these changes. To address this problem, continual learning techniques for KGEs incorporate embeddings for new entities while updating the old ones. One necessary step in these methods is the initialization of the embeddings, as an input to the KGE learning process, which can have an important impact in the accuracy of the final embeddings, as well as in the time required to train them. This is especially relevant for relatively small and frequent updates. We propose a novel informed embedding initialization strategy, which can be seamlessly integrated into existing continual learning methods for KGE, that enhances the acquisition of new knowledge while reducing catastrophic forgetting. Specifically, the KG schema and the previously learned embeddings are utilized to obtain initial representations for the new entities, based on the classes the entities belong to. Our extensive experimental analysis shows that the proposed initialization strategy improves the predictive performance of the resulting KGEs, while also enhancing knowledge retention. Furthermore, our approach accelerates knowledge acquisition, reducing the number of epochs, and therefore time, required to incrementally learn new embeddings. Finally, its benefits across various types of KGE learning models are demonstrated.

Denial Constraints (DCs) are a flexible formalism to express many types of data rules, making them a widely adopted tool for many applications. This flexibility led to the development of numerous algorithms to automatically discover DCs directly from data. However, few studies have been conducted on the quality of the discovered DCs. We experimentally quantify the lack of quality in the results obtained by state-of-the-art algorithms, showing how the proportion of discovered DCs that are false is rarely below 95%. We hypothesize that the common source of these erroneous DCs stems from the adoption of the current DC validity definition. We use a statistical approach to explain the mechanism leading to these results, and propose a redefinition of DC validity properties to avoid the acceptance of false DCs. We validate this redefinition experimentally, showing that it exclusively accepts true constraints of the data, and is reliable enough to discover DCs missed by domain experts. Additionally, we provide curated sets of golden DCs for each dataset used in our study, those generated by domain experts and those discovered using our approach.

The popularity of heterogeneous CPU-GPU processing has increased considerably in recent years. To efficiently utilize heterogeneous resources, data processing systems depend on an appropriate workload placement strategy to assign the right amount of compute to the right processor. However, finding an optimal placement strategy is not trivial due to various complex and conflicting tradeoffs related to the characteristics of processors, the nature of the workload, and data locality. In addition, placement decisions impact workload runtime and performance cost, and also depend on the availability of potentially different implementations for CPUs and GPUs, which adds extra complexity in such heterogeneous environments. In this tutorial, we review and compare state-of-the-art strategies for workload placement on heterogeneous CPU-GPU architectures, along with runtime prediction techniques and methods to support multi-device code. We also discuss open issues and identify potentially promising future research directions.

Recent advances in Large Language Models (LLMs) have positioned them as a prominent solution for Natural Language Processing tasks. Notably, they can approach these problems in a zero or few-shot manner, thereby eliminating the need for training or fine-tuning task-specific models. However, LLMs face some challenges, including hallucination and the presence of outdated knowledge or missing information from specific domains in the training data. These problems cannot be easily solved by retraining the models with new data as it is a time-consuming and expensive process. To mitigate these issues, Knowledge Graphs (KGs) have been proposed as a structured external source of information to enrich LLMs. With this idea, in this work we use KGs to enhance LLMs for zero-shot Entity Disambiguation (ED). For that purpose, we leverage the hierarchical representation of the entities' classes in a KG to gradually prune the candidate space as well as the entities' descriptions to enrich the input prompt with additional factual knowledge. Our evaluation on popular ED datasets shows that the proposed method outperforms non-enhanced and description-only enhanced LLMs, and has a higher degree of adaptability than task-specific models. Furthermore, we conduct an error analysis and discuss the impact of the leveraged KG's semantic expressivity on the ED performance.

This article introduces a methodology for a novel data-driven computational model aimed at aiding public administrations in managing and evaluating the adaptative reuse of buildings while tackling ecological and digital challenges. Drawing from the 2030 Agenda for Sustainable Development, the study underscores the significance of innovative approaches in harnessing the economic potential of data. Focusing on Barcelona’s Ciutat Vella district, the research selects five historic public buildings for analysis, strategically positioned to spur local entrepreneurship and counteract tourism dominance. Through an extensive literature review, the article identifies a gap in computational models for building adaptative reuse and proposes a methodological framework that integrates data collection, processing, and computational modelling, underscored by GIS technology and open data sources. The proposed methodology for a computational algorithm aims to systematise spatial characteristics, assess programmatic needs, and optimise building usage, while addressing challenges such as data integration and quality assurance. Ultimately, the research presents a pioneering approach to building adaptative reuse, aimed at fostering sustainable urban development and offering replicable insights applicable to similar challenges in other cities.

Task-based programming models have proven to be a robust and versatile way to approach development of applications for distributed environments. They provide natural programming patterns with high performance. However, execution on this paradigm can be very sensitive to granularity –i.e., the quantity and execution length of tasks. Granularity is often linked with the block size of the data, and finding the optimal block size has several challenges, as it requires inner knowledge of the computing environment. Our proposal is to supplement the task-based programming model with a new mechanism –our SplIter proposal. At its core, the SplIter provides a transparent way to split a collection into partitions (logical groups of blocks, obtained without any transfers nor data rearrangement), which can then be iterated. Tasks are linked to those partitions, which means that SplIter breaks the dependency between block size and task granularity. The evaluation shows that the SplIter is able to achieve performance improvements of over one order of magnitude when compared to the baseline, and it is either competitive or strictly better (depending on application characteristics) to the competitor alternative. We have chosen different applications covering a wide variety of scenarios; those applications are representatives of a broader set of applications and domains. The changes required in the source code of a task-based application are minimal, preserving the high programmability of the programming model. Two different state-of-the-art task-based frameworks have been evaluated for all the applications: COMPSs and Dask, showing that the SplIter can be effectively used within different frameworks.

Object stores are widely used software stacks that achieve excellent scale-out with a well-defined interface and robust performance. However, their traditional get/put interface is unable to exploit data locality at its fullest, and limits reaching its peak performance. In particular, there is one way to improve data locality that has not yet achieved mainstream adoption: the active object store. Although there are some projects that have implemented the main idea of the active object store such as Swift's Storlets or Ceph Object Classes, the scope of these implementations is limited. We believe that there is a huge potential for active object stores in the current status quo. Hyper converged nodes are bringing more computing capabilities to storage nodes - and vice versa. The proliferation of non-volatile memory (NVM) technology is blurring the line between system memory (fast and scarce) and block devices (slow and abundant). More and more applications need to manage a sheer amount of data (data analytics, Big Data, Machine Learning & AI, etc.), demanding bigger clusters and more complex computations. All these elements are potential game changers that need to be evaluated in the scope of active object stores. More specifically, having NVM devices presents additional opportunities, such as in-place execution. Being able to use the NVM from within the storage system while taking advantage of in-place execution (thanks to the byte-addressable nature of the NVM), in conjunction with the computing capabilities of hyper-converged nodes, can lead to active object stores that greatly outperform their non-active counterparts. In this article we propose an active object store software stack and evaluate it on an NVM-populated node. We will show how this setup is able to reduce execution times from 10% up to more than 90% in a variety of representative application scenarios. Our discussion will focus on the active aspect of the system as well as on the implications of the memory configuration. (c) 2021 Elsevier B.V. All rights reserved.