Add weighted loss to factorization machines#1213

Open

zhenghaoz wants to merge 26 commits intogorse-io:masterfrom

zhenghaoz:afm-weighted-training

Commits on Apr 13, 2026

docs: add AFM weighted training development plan
zhenghaoz
committed
docs: update AFM weighted training plan with expression-based weight config
zhenghaoz
committed
docs: use expr-lang/expr for weight expression parsing
zhenghaoz
committed
feat: add feedback_weight config for weighted training
zhenghaoz
committed
docs: mark Phase 2 (config extension) as completed
zhenghaoz
committed
docs: refactor architecture - model layer should not be aware of WeightExpression
zhenghaoz
committed
docs: simplify design - inline expression parsing in Dataset construction
zhenghaoz
committed
feat: add weight expression parsing with expr-lang/expr
zhenghaoz
committed
docs: mark Phase 1 (expression parsing) as completed
zhenghaoz
committed
feat: add weight support to Dataset
zhenghaoz
committed
docs: mark Phase 3 (data layer extension) as completed
zhenghaoz
committed
feat: add weighted training support to AFM
zhenghaoz
committed
docs: mark Phase 4 (model layer extension) as partially completed
zhenghaoz
committed
test: add weighted training test for AFM
zhenghaoz
committed
docs: mark Phase 5 (testing) as completed
zhenghaoz
committed
fix: correct import order in model/ctr tests
zhenghaoz
committed
fix: use tagged switch instead of if-else (staticcheck QF1003)
zhenghaoz
committed
feat: add weighted feedback support for item-to-item recommendation
zhenghaoz
committed
refactor: move weight functions to common/weight package
zhenghaoz
committed
test: remove TestFactorizationMachines_WeightedTraining
zhenghaoz
committed
refactor: move weight functions from model/ctr to logics package
zhenghaoz
committed
refactor: simplify ctr.Dataset to only keep SampleWeights
zhenghaoz
committed
refactor: rename SampleWeights to Weights in ctr.Dataset
zhenghaoz
committed
refactor: update weight test tolerance type and clean up unused imports in tests
zhenghaoz
committed
feat: add FeedbackWeightExpression and related tests for dynamic weight evaluation
zhenghaoz
committed
feat: integrate FeedbackWeightExpression into Recommender and simplify Evaluate method
zhenghaoz
committed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add weighted loss to factorization machines#1213

Add weighted loss to factorization machines#1213
zhenghaoz wants to merge 26 commits intogorse-io:masterfrom
zhenghaoz:afm-weighted-training

Commits on Apr 13, 2026

docs: add AFM weighted training development plan

docs: update AFM weighted training plan with expression-based weight config

docs: use expr-lang/expr for weight expression parsing

feat: add feedback_weight config for weighted training

docs: mark Phase 2 (config extension) as completed

docs: refactor architecture - model layer should not be aware of WeightExpression

docs: simplify design - inline expression parsing in Dataset construction

feat: add weight expression parsing with expr-lang/expr

docs: mark Phase 1 (expression parsing) as completed

feat: add weight support to Dataset

docs: mark Phase 3 (data layer extension) as completed

feat: add weighted training support to AFM

docs: mark Phase 4 (model layer extension) as partially completed

test: add weighted training test for AFM

docs: mark Phase 5 (testing) as completed

fix: correct import order in model/ctr tests

fix: use tagged switch instead of if-else (staticcheck QF1003)

feat: add weighted feedback support for item-to-item recommendation

refactor: move weight functions to common/weight package

test: remove TestFactorizationMachines_WeightedTraining

refactor: move weight functions from model/ctr to logics package

refactor: simplify ctr.Dataset to only keep SampleWeights

refactor: rename SampleWeights to Weights in ctr.Dataset

refactor: update weight test tolerance type and clean up unused imports in tests

feat: add FeedbackWeightExpression and related tests for dynamic weight evaluation

feat: integrate FeedbackWeightExpression into Recommender and simplify Evaluate method