Skip to content
KortexyaKortexyaReasoning Layer SDK

Interface: LtnTrainResponse

@kortexya/reasoninglayer

@kortexya/reasoninglayer / LtnTrainResponse

Interface: LtnTrainResponse

Section titled “Interface: LtnTrainResponse”

Defined in: src/types/ltn.ts:433

The result of training: the SatAgg trace and learned certainties, plus an optional neural-predicate fit summary.

Example

Section titled “Example”
console.log(response.initialSat, '→', response.finalSat);
for (const learned of response.learnedCertainties) {
  console.log(learned.individual, learned.initialCertainty, '→', learned.certainty);
}

Properties

Section titled “Properties”

epochsRun

Section titled “epochsRun”

epochsRun: number

Defined in: src/types/ltn.ts:435

Number of certainty-training epochs actually run.

finalSat

Section titled “finalSat”

finalSat: number

Defined in: src/types/ltn.ts:439

SatAgg after the final epoch.

initialSat

Section titled “initialSat”

initialSat: number

Defined in: src/types/ltn.ts:437

SatAgg before training (epoch 0).

learnedCertainties

Section titled “learnedCertainties”

learnedCertainties: LtnLearnedCertainty[]

Defined in: src/types/ltn.ts:443

Learned certainty for each rule, in request order.

perEpochSat

Section titled “perEpochSat”

perEpochSat: number[]

Defined in: src/types/ltn.ts:441

SatAgg after each epoch (early-stop diagnostics).

predicate?

Section titled “predicate?”

optional predicate?: LtnPredicateFit | null

Defined in: src/types/ltn.ts:445

Neural-predicate fit summary, when a predicate block was supplied.