Shear MomentsML Training

Data product name

sheMomentsMlTraining

Data product custodian

SHE

Data model tag

8.0

Name of the Schema file

ST_DM_Schema/dictionary/pro/she/euc-test-she.xsd

Schema documentation tag

Documentation for data product element sheMomentsMlTraining:

This product contains the trained parameters of the MomentsML shear estimation method.

Documentation for data product element Header:

The generic header of the product.

Data product elements

Base class: cat:baseCatalog

• ValidityRange (ins:calibrationValidPeriod)

• DataStorage (sheGenericFile)

• DataContainer (dss:dataContainer)

Processing Element(s) creating/using the data product

SHE

Processing function using the data product

SHE

Detailed description of the data product

An XML data product containing trained parameters of the MomentsML shear measurement algorithm. The data consists mainly of parameters (weights and biases) of artificial neural networks trained to predict shear, and of a description of the architecture of these networks. Given the complex structure of this data, it is serialized into an binary file, in a format supported by the TensorFlow and Keras libraries.

For MomentsML, the distinction between “training” and “calibration” is that training refers to the optimization of the machine-learning model itself, while calibration offers a possibility to address effects not captured by the training simulations of the machine learning.