Conference Agenda

Determining the Quantified Semantic Similarity (QSS) between database queries is a critical challenge with broad applications, from query log analysis to automated SQL skill assessment. Traditional methods often rely solely on syntactic comparisons or are limited to checking for semantic equivalence.

This paper introduces Graph-based QSS, a novel graph-based approach to measure the semantic dissimilarity between SQL queries. Queries are represented as nodes in an implicit graph, while the transitions between nodes are called edits, which are weighted by semantic dissimilarity. We employ shortest path algorithms to identify the lowest-cost edit sequence between two given queries, thereby defining a quantifiable measure of semantic distance.

An empirical study of our prototype suggests that our method provides more accurate and comprehensible grading compared to existing techniques. Moreover, the results indicate that our approach comes close to the quality of manual grading, making it a robust tool for diverse database query comparison tasks.

Spatial data is generated daily from numerous sources such as GPS-enabled devices, consumer applications (e.g., Uber, Strava), and social media (e.g., location-tagged posts). This exponential growth in spatial data is driving the development of efficient spatial data processing systems.

In this study, we enhance spatial indexing with a machine-learned search technique developed for single-dimensional sorted data. Specifically, we partition spatial data using six traditional spatial partitioning techniques and employ machine-learned search within each partition to support point, range, distance, and spatial join queries. By instance-optimizing each partitioning technique, we demonstrate that:

(i) grid-based index structures outperform tree-based ones (from 1.23x to 2.47x),

(ii) learning-enhanced spatial index structures are faster than their original counterparts (from 1.44x to 53.34x),

(iii) machine-learned search within a partition is 11.79% - 39.51% faster than binary search when filtering on one dimension,

(iv) the benefit of machine-learned search decreases in the presence of other compute-intensive operations (e.g. scan costs in higher selectivity queries, Haversine distance computation, and point-in-polygon tests), and

(v) index lookup is the bottleneck for tree-based structures, which could be mitigated by linearizing the indexed partitions.

Schema matching involves identifying matching relationships between different data schemas. While this task is supported by semi-automatic tools, achieving optimal results requires configuring such tools which can be challenging depending on the number of configuration options and schema characteristics. This study proposes a novel approach utilizing Reinforcement Learning (RL) to automate the configuration of schema matching tools. RL has proven to be well-suited for complex optimization problems but has not yet been applied for schema matching. We outline how the configuration of a schema matching tool can be tackled as an RL task and how the corresponding learning process can be accelerated and optimized. We evaluate the RL approach for a large real-world dataset and show that it can be applied to different matching tools.

The growing field of explainable AI (XAI) develops methods that help better understand ML

model predictions. While SHapley Additive exPlanations (SHAP) is a widely-used, model-agnostic method for explaining predictions, its use comes with a significant computational burden, particularly for complex models and large datasets with many features. The key—and so far unaddressed—challenge lies in efficiently scaling these computations without compromising accuracy. In this paper, we present a scalable, model-agnostic SHAP sampling framework on top of Apache SystemDS. We leverage Antithetic Permutation Sampling for its efficiency and optimization potential, and we devise a carefully vectorized and parallelized implementation for local and distributed operations. Compared with the state-of-the-art Python shap package, our solutions yield similar accuracy but achieve significant speedups of up to 14x for multi-threaded singlenode operations as well as up to 35x for distributed Spark operations (on a small 8 node cluster).

Clustering is a fundamental data analytics operation in the field of unsupervised learning. Given a database of unknown structure, clustering aims to discover the inherent structure of the data objects according to similarity, such that similar data objects are grouped together, while dissimilar ones are separated in different groups or clusters.

In this study, we focus on partitioning methods, specifically the k-means clustering approach, which minimizes the intra-cluster variance.

As the standard algorithmic approach to k-means clustering, namely the Lloyd algorithm, is neither efficient nor scalable, various adaptations and modifications have been developed, resulting in the family of fast k-means clustering algorithms. In this short paper, we extend the clustering algorithm proposed by Elkan with Ptolemy's inequality to prune superfluous distance calculations. It is not our intention with this paper to compete with the state-of-the-art algorithmic solutions for the k-means clustering problem; rather, we seek to investigate the potential of Ptolemy’s inequality to further enhancements in the standard algorithm’s performance, particularly in terms of reducing computations associated with distance evaluations.