LOFAR Dataset for Deep Learning Assisted Radio Astronomy Data Inspection

A dataset used for training magnitude and phase-based Variational Autoencoders (VAEs) in the paper "Deep learning assisted data inspection for radio astronomy". The data originates from the LOFAR radio telescope and is provided by Michael Mesarcik of the University of Amsterdam. The dataset is distributed as four separate zip files that must be extracted into a single directory for training.

Use Cases

• Training magnitude-based Variational Autoencoders for anomaly detection in radio astronomy data.
• Training phase-based Variational Autoencoders for feature extraction from radio telescope signals.
• Developing deep learning models for automated data inspection and quality control in astronomical observations.

Strengths

• Dataset is explicitly designed for training both magnitude and phase-based deep learning models.
• Associated with a published academic paper, providing a documented research context.
• Data originates from the LOFAR radio telescope, a major astronomical instrument.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

LOFAR radio telescope, University of Amsterdam

Collection Method

Data collected by the LOFAR instrument, processed for machine learning training.