Dontopedia

Straight Through Surrogate

From Dontopedia, the open, paraconsistent wiki. (Last updated 2026-06-06.)

Straight Through Surrogate has 6 facts recorded in Dontopedia across 3 references.

6 facts·6 predicates·3 sources

Mostly:hinders gradients(1), approximates(1), compared to(1)

Maturity scale raw canonical shape-checked rule-derived certified

Inbound mentions (4)

Other subjects in dontopedia point AT this entity as a value. These are inverse relationships — e.g. "X motherOf this subject" — and answer questions the forward facts can't. Grouped by predicate.

dropsTechniqueDrops Technique(1)

lacksStraightThroughSurrogateLacks Straight Through Surrogate(1)

methodMethod(1)

usesUses(1)

Other facts (6)

The long tail: predicates that appear too rarely to warrant their own section. Filter or scroll to find a specific one. Each row links to its source.

6 facts
PredicateValueRef
Hinders GradientsPer Group Scalars[1]
ApproximatesProper Bptt[2]
Compared toProper Bptt Compute Cost[2]
Compute Cost Percentage of0.9[2]
Rdf:typeMethod[3]
Compute Cost Relation90% of proper BPTT compute cost[3]

Timeline

Timeline axis is valid_time — when each source says the fact was true in the world, not when Dontopedia learned about it. Retracted rows are kept for provenance; coloured stripes indicate the context kind.

hindersGradientsblah/watt-activation/part-634
ex:per-group-scalars
approximatesblah/watt-activation/part-640
ex:proper-bptt
comparedToblah/watt-activation/part-640
ex:proper-bptt-compute-cost
computeCostPercentageOfblah/watt-activation/part-640
0.9
typeblah/watt-activation/637
ex:Method
computeCostRelationblah/watt-activation/637
90% of proper BPTT compute cost

References (3)

3 references
  1. [1]Part 6341 fact
    ctx:discord/blah/watt-activation/part-634
  2. [2]Part 6403 facts
    ctx:discord/blah/watt-activation/part-640
  3. [3]6372 facts
    ctx:discord/blah/watt-activation/637
    • full textwatt-activation-637
      text/plain3 KBdoc:agent/watt-activation-637/19b1b41c-2a51-4202-8887-489a5512853c
      Show excerpt
      [2026-04-16 00:35] xenonfun: Proposed new order 1. Batching fix for training path (quick) — apply the same batched_linear_forward change to forward_train and its backward. Free 2.5× on training throughput. This must happen before Optio

See also

Keep researching

Missing something or suspicious of what's here? Kick off a research session — a Claude agent will investigate, cite its sources, and file new facts into a dedicated context you can review before accepting into the shared view.