How mitchellkeegan/accel-oct works
This is research software for training optimal classification trees with integer programming and producing the paper’s experimental results. The main workflow is to configure and run an OCT model, prepare encoded data, build and optimize a MIP, then log and post-process outcomes into tables and figures. The graph preserves the documented model families and acceleration options, including cuts, callbacks, EQP sets, and solution polishing. Some subsystem links are limited to README statements and SOURCE INDEX references; unsampled implementation details are not inferred.
Main entry points: base_classes.py OCT framework · RunOCT.py Run interface · BendOCT.py Bend OCT · BendRegOCT.py Regularized Bend
Components 22
Experiment entry
- Run interface
Scripts/RunOCT.py - Dictionary interface
Scripts/OCTDictInterface.py
OCT models
- Bend OCT
src/models/BendOCT.py - Regularized Bend
src/models/BendRegOCT.py - Worst-case Bend
src/models/BendWorstOCT.py - Flow OCT
src/models/FlowOCT.py - Regularized Flow
src/models/FlowRegOCT.py - OCT framework
src/models/base_classes.py
Model acceleration
- Cuts and callbacks
src/models/BendOCT.py - Regularized cuts and callbacks
src/models/BendRegOCT.py - Path-bound callbacks
src/models/BendWorstOCT.py - EQP set generation
src/utils/generators.py
Data and runtime support
- Dataset loading
src/utils/data.py - Tree routines
src/utils/trees.py - Pareto filtering
src/utils/ParetoSets.py - Run logging
src/utils/logging.py
Result analysis
- Result processing
Scripts/ResultsProcessing/CreateTables.py - Aggregate tables
Scripts/ResultsProcessing/CreateAggTable.py - Benchmark figures
Scripts/ResultsProcessing/BenchmarkFigure.py - Benchmark tables
Scripts/ResultsProcessing/BenchmarkTable.py
Other components
- Researcher
- MIP solver
Connections 35
- Researcher → Run interface: runs
- Run interface → Dictionary interface: wraps
- Run interface → Dataset loading: loads data
Scripts/RunOCT.py - Run interface → Bend OCT: configures
Scripts/RunOCT.py - Run interface → Regularized Bend: configures
Scripts/RunOCT.py - Run interface → Regularized Flow: configures
Scripts/RunOCT.py - Dictionary interface → Bend OCT: selects model
Scripts/OCTDictInterface.py - Dictionary interface → Regularized Bend: selects model
Scripts/OCTDictInterface.py - Dictionary interface → Regularized Flow: selects model
Scripts/OCTDictInterface.py - Bend OCT → OCT framework: inherits
src/models/BendOCT.py - Regularized Bend → OCT framework: inherits
src/models/BendRegOCT.py - Worst-case Bend → OCT framework: inherits
src/models/BendWorstOCT.py - Flow OCT → OCT framework: inherits
src/models/FlowOCT.py - Regularized Flow → OCT framework: inherits
src/models/FlowRegOCT.py - OCT framework → MIP solver: optimizes
- OCT framework → Cuts and callbacks: configures
src/models/BendOCT.py - OCT framework → Regularized cuts and callbacks: configures
src/models/BendRegOCT.py - OCT framework → Path-bound callbacks: configures
src/models/BendWorstOCT.py - Bend OCT → Tree routines: uses
src/models/BendOCT.py - Bend OCT → EQP set generation: uses
src/models/BendOCT.py - Regularized Bend → Tree routines: uses
src/models/BendRegOCT.py - Regularized Bend → EQP set generation: uses
src/models/BendRegOCT.py - Worst-case Bend → Tree routines: uses
src/models/BendWorstOCT.py - Flow OCT → Tree routines: uses
src/models/FlowOCT.py - Regularized Flow → Tree routines: uses
src/models/FlowRegOCT.py - EQP set generation → Pareto filtering: filters sets
- Bend OCT → Run logging: logs
src/models/BendOCT.py - Regularized Bend → Run logging: logs
src/models/BendRegOCT.py - Worst-case Bend → Run logging: logs
src/models/BendWorstOCT.py - Flow OCT → Run logging: logs
src/models/FlowOCT.py - Regularized Flow → Run logging: logs
src/models/FlowRegOCT.py - Result processing → Dataset loading: reads
Scripts/ResultsProcessing/CreateTables.py - Aggregate tables → Dataset loading: reads
Scripts/ResultsProcessing/CreateAggTable.py - Benchmark figures → Dataset loading: reads
Scripts/ResultsProcessing/BenchmarkFigure.py - Benchmark tables → Dataset loading: reads
Scripts/ResultsProcessing/BenchmarkTable.py