Inhalt des Dokuments
Es gibt keine deutsche Übersetzung dieser Webseite.
Paper Accepted for the Gongshow Presentation at CIDR 2020
The paper
"CAFE: Constraint-Aware Feature
Extraction from Large Databases"
by Mahdi Esmailoghli and Ziawasch Abedjan was accepted for the Gongshow Presentation at CIDR 2020.
Here is the abstract:
"Feature extraction is a core step of most
machine learning pipelines. If a given dataset is not providing useful
features one has to enhance the dataset at hand with features from
additional sources. This will turn into a manual and tedious search
and retrieval process.
At the same time, users may want to uphold certain constraints on the new features, such as consistency of values, interpretability, or fairness.
We propose a constraint-aware feature extraction method that scans a given large database for promising features and effectively translates user-defined constraints into extraction filters. To this end, we show how the initial retrieval engine of our system can swift through million of tables to find relevant features and how our execution planner optimizes the required and data-dependent pruning process for fast and effective feature extraction."
Zusatzinformationen / Extras
Direktzugang:
Schnellnavigation zur Seite über Nummerneingabe
Hilfsfunktionen
- Schriftgrößee:
Copyright TU Berlin 2008