Datenbank-Spektrum - Call-for-Papers: ML for Systems and Systems for ML
Machine learning (ML) methods are data-intensive, and data engineering workflows are an essential part of all machine learning workflows. They provide data for the specific machine learning tasks and usually combine multiple data engineering tasks. Within the lifecycle of a machine learning pipeline, collecting, storing, and selecting the data for (re)training the model and deploying the model (in a distributed environment) are performance-intensive parts. In this way, the data engineering tasks enable machine learning solutions.
It is also possible to use machine learning methods for individual data engineering tasks (like data exploration, discovery, certain data cleaning tasks, and data integration) or to optimize the data engineering processes. Machine learning for systems aims to replace components of systems with learned models. Database systems benefit from learned components such as cardinality estimation or query optimizer hinting.
Another current research topic is the development of systems for machine learning to support design of end-to-end machine learning solutions. For data engineering, systems are fundamental for processing data and their enhancement is crucial for efficient model training and inference. Thus, systems for machine learning enhance the application of models by integrating machine learning functionality into data management systems.
In this way, the fields of databases and machine learning are growing closer together and are mutually dependent. We want this special issue to highlight the resulting new approaches.
We welcome original contributions or extensions of previous work on the following topics (but not limited to):
- Learned system components or algorithms (such as index structures, query optimization and optimizer hinting)
- Self-tuning of database systems
- Data cleaning and monitoring techniques for ML workflows
- Data quality management and evaluation for ML workflows
- ML methods for data exploration, discovery, and integration
- Data, metadata and model management in ML applications and complex ML pipelines
- Novel use of natural language models and interfaces in data management
- Novel use of ML techniques in big data applications (e.g., digital biology, video analytics)
- New datasets, benchmarks, and evaluation methods for evaluating ML approaches
- Data integration, alignment, and preparation of multi-modal training datasets
- New/Extended Systems for ML (Database Systems extended for ML, ML systems extended by compression/indexing/partitioning)
- Physical/Logical optimization of ML pipelines (data flow by operator rewriting/fusion, data-/task-parallel execution strategies)
- Further topics combining database and machine learning technologies
Paper format: 8–10 pages, double-column (author guidelines at http://www.www.springer.com/13222 (this opens in a new tab) ). We welcome contributions in both, German, and English.
Articles will undergo all the journal's standard peer review and editorial processes outlined in its submission guidelines (this opens in a new tab) .
Deadline for submissions: June 1st, 2024
Publication of special issue: DASP-3-2024 (November 2024)
Guest editors:
Maximilian E. Schüle, University of Bamberg, Germany, maximilian.schuele@uni-bamberg.de (this opens in a new tab)
Meike Klettke, University of Regensburg, Germany, meike.klettke@ur.de (this opens in a new tab)