assistant= TS + β·log(1+A) - γ·V---I found it in a paper about ML training runs and I want to know what each term means