Question about MAT_CONST_H_MAX in paderborn.py

[moeKedama]

MAT_CONST_H_MAX = {
"3C90": 84.7148502254261,
"3C94": 64.8575649838852,
"3E6": 74.1579701817075,
"3F4": 150,
"77": 86.5681744566843,
"78": 87.5896894086919,
"3C92": 150,
"T37": 87.7490689795367,
"3C95": 72.0625845199264,
"79": 150,
"ML95S": 150,
"N27": 119.039616554254,
"N30": 116.951204964406,
"N49": 150,
"N87": 100.674197407678,}  # in A/m    

I noticed that the MAT_CONST_H_MAX of some materials is set to 150, but their actual values may exceed this number. I don't quite understand the purpose of this operation, could someone provide an explanation, please? Thank you.

[wkirgsn]

Hi, these values denote the denominator during a simple scaling operation that represents the normalization of the data.
NN training benefits from normalized data.
Some materials are clipped to 150 A/m as we found during our analyses that the maximum occurring H values for these materials rather resemble outliers and would scale the entirety of the data inappropriately.

Hope I could clarify these values ✌🏼