Manuscript Psy03-A@1 · Psy03-A.tex · section · b30eb6e9

The source metrics

Intervention budgets and the failure of weight-only dose › The source metrics · lines 381–422

When masked parameters have been enforced, the source's primary weight-displacement quantity is

D_{W} = \frac{\norm vec ( B ^{'} - B ) _{1}}{N _{W}} .

Its reported “L2 dose” is

D_{2, code} = \frac{\norm vec ( B ^{'} - B ) _{2}}{N _{W}},

not the root-mean-square displacement

D_{2, RMS} = \frac{\norm vec ( B ^{'} - B ) _{2}}{N _{W}} .

Consequently,

D_{2, code} \leq D_{W} \leq N_{W} D_{2, code}, D_{2, RMS} = N_{W} D_{2, code} .

These identities follow from the finite-dimensional norm inequalities.

The relative-change statistic

T_{rel} = \frac{1}{N _{W}} i \sum 1 {\frac{\abs B _{i}^{'} - B _{i}}{max ( \abs B _{i} , 1 0 ^{- 8} )} > 0.10}

is weight turnover, not structural turnover. At a baseline zero position its threshold is $1 0^{- 9}$ . It does not distinguish a mask birth from a change of an already active weight. Numerical zero counts and mask counts are different as well.

Displayed formulas in this passage

unlabeled · line 385
$D_{W} = \frac{\norm vec ( B ^{'} - B ) _{1}}{N _{W}} .$
unlabeled · line 389
$D_{2, code} = \frac{\norm vec ( B ^{'} - B ) _{2}}{N _{W}},$
unlabeled · line 394
$D_{2, RMS} = \frac{\norm vec ( B ^{'} - B ) _{2}}{N _{W}} .$
unlabeled · line 399
$D_{2, code} \leq D_{W} \leq N_{W} D_{2, code}, D_{2, RMS} = N_{W} D_{2, code} .$
unlabeled · line 409
$T_{rel} = \frac{1}{N _{W}} i \sum 1 {\frac{\abs B _{i}^{'} - B _{i}}{max ( \abs B _{i} , 1 0 ^{- 8} )} > 0.10}$

The source metrics

Displayed formulas in this passage

Nearby in this section