Machine Learning Models Product

Metadata

Data product name

DpdMerMachineLearningModels

Data product custodian

MER

Name of the Schema file

euc-mer-MachineLearningModels.xsd

Last Edited for DPDD Version

2.0

Processing Element(s) creating/using the data product

Creators:

  • This is an input product that is generated manually or using some dedicated

scripts.

Consumers:

Processing function using the data product

MER

Proposed for inclusion in EAS/SAS

This product is proposed for inclusion in the SAS: maybe.

This product contains trained machine learning models that are used in different steps of the MER pipeline. It could be useful to have them in the SAS in case scientists want to replicate some of the MER classifications.

Data Product Elements

Header

object of type sys:genericHeader

Data

object of type mer:merMachineLearningModels

QualityFlags

object of type dqc:sqfPlaceHolder

Parameters

object of type ppr:genericKeyValueParameters

Detailed Description of the Data Product

This product is an input to the MER Pipeline. It contains a list of trained machine learning (ML) models used in one or several of the MER processing steps.

The product currently contains a ML model trained to detect spurious sources, and the Zoobot model used to classify sources according to their morphology (see the MER morphology cookbook for more details).

The main elements inside this product are:

  • TileIndex (optional): The MER tile index to which the machine learning models should be associated.

  • PatchId (optional): The sky patch id to which the machine learning models should be associated.

  • CalblockId (optional): The calibration block id to which the machine learning models should be associated.

  • CalblockVariant (optional): The calibration block variant to which the machine learning models should be associated.

  • ProcessingMode (optional): The MER pipeline processing mode (e.g. WIDE, DEEP) to which the machine learning models should be associated.

  • Model (one to infinity): The model name and the file containing the trained model.