Read the equation
The bend statistic is evaluated with lags 5, 10, and 15 on each side of a center checkpoint. Larger lags compare broader segments of the path. These are sensitivity settings applied to the same training run, rather than separate runs.
Terms & definitions
| Term | Meaning in this source |
|---|---|
| k | Two-sided checkpoint lag |
How it works
Read the definition from its right-hand expression to the quantity on the left. Its surrounding source records the intended input, state, or parameter context. Where the source leaves an operation unspecified, this page preserves that gap.
Assumptions & limits
Source recovery establishes that the expression appears in this record. Its mathematical validity, implementation behavior, and interpretation must be assessed separately. See the linked source and research notes for the checks available for this version.
Original context
Read the surrounding source passage
The local metric used in the released analysis compares displacements from a center checkpoint to earlier and later checkpoints [3]. For Zt=bt or Zt=Ut, define
cZt;k=⟨Zt−k−Zt,Zt+k−Zt⟩∥Zt−k−Zt∥ ∥Zt+k−Zt∥.
This is not the cosine between Zt and Zt+k. A straight, continuing path gives cZ=−1, since the backward and forward displacements point oppositely. Larger values indicate a sharper local bend. The complete windows are evaluated on five-step checkpoint centers for lags k∈{5,10,15}. Each lag uses every center for which its own two neighbors exist; this need not reproduce identical edge trimming in every source plotting call.
An angle can emphasize a tiny late-training movement. We therefore also apply the source program’s factor-amplitude eligibility rule. Let mA be the larger of the preceding and succeeding input-factor displacement norms, and mB the corresponding output-factor quantity. A center is excluded only when both are below threshold η. The sweep is η∈{0,10−4,5×10−4,10−3,2×10−3,5×10−3}.
[Context excerpt; full source linked.]Exact original notation
k\in {5,10,15}