hugiml-core Logo

User guide

  • Getting started
    • Installation
    • Minimal classifier workflow
    • Cross-validation and production schemas
    • Downstream solver choices
    • Recommended first checks
    • Recommended starting points
  • Core concepts
    • HUG-IML process
    • Pattern anatomy
    • Key hyperparameters
    • Missing values
    • Adaptive binning
    • Performance and mining behavior
    • Operational stability controls
    • Constant and zero-utility columns
  • Feature modes
    • Recommended configurations
    • No-pattern behavior
    • Augmented pair features
    • Interaction-relaxed mining
    • Examples
    • Downstream matrix policy
    • Interpretation notes
    • Compatibility with transform
    • Operational guidance
  • Execution modes
    • Basic usage
    • Production-mode guidance
    • Recommended workflow
    • Fallback state
  • Augmented pair features
    • When they are active
    • Source selection modes
    • Feature families
    • Example
    • Interpretation
    • Interaction-relaxed alternative
    • Budgeting
    • Review workflow
  • RPTE downstream model
    • Overview
    • Conceptual model
    • How tree growth works
    • Boosting and final sparse layer
    • Interpretability
    • Basic use
    • Binary and multiclass behavior
    • Configuration guidance
  • Complexity units
    • Model units
    • Model inspection units
    • Instance inspection units
  • Fast hyperparameter tuning
    • Recommended named grids
    • Cross-validated tuning
    • Focused follow-up grids
    • Fast path eligibility
    • Computational budget notes
  • Native mining pruning controls
    • Pruning paths
    • Native hot paths
    • Mining timeout and audit log
    • EUCS parameters
    • Practical guidance
  • Explanations and dashboards
    • Pattern tables
    • Hybrid feature interpretation
    • RPTE explanations
    • Interpretability metrics
    • Plotly dashboard
    • Profile views
    • HUGIML versus additive shape models
    • SHAP bridge
  • Governance Studio dashboard
    • Installation and launch
    • Workbench workflow
    • Data profiling
    • Evidence views
    • Data inputs
    • Binary feature handling
    • Installed command behavior
    • Dashboard API modules
  • LLM assistant add-on
    • Installation
    • Fast and thinking response modes
    • Supported workflows
    • Built-in assets
    • API modules
  • Governance, pruning, and model cards
    • Governance Studio evidence
    • Binary feature evidence
    • Model cards
    • Pattern pruning
    • Audit artifacts
    • Calibration
  • Monitoring, drift, and serialization
    • Prediction monitoring
    • Drift detection
    • Serialization
    • Telemetry
  • Deployment
    • FastAPI inference server
    • Kubernetes
    • Production checklist
  • Benchmarks
    • OpenML-CC18 benchmark
    • Compared models
    • Predefined datasets
    • Custom datasets
    • Tuned benchmarks
    • Programmatic use
    • Outputs
    • Benchmark visuals
    • Missing-value robustness
    • Interpretation guidance
    • Reproducibility notes
  • Worked notebooks
    • Publishing notebooks
  • Publishing on Read the Docs
    • Repository files used by Read the Docs
    • Recommended update workflow
    • Updating the existing hugiml-core RTD page
    • Local build check
    • PyPI and docs alignment
    • Troubleshooting
  • Paper and scientific scope
    • Reference
    • Scope reflected in this implementation
    • Citation

API reference

  • API reference
    • Core estimator
      • Architecture
      • Quick start
      • HUGIMLClassifier
      • HUGIMLClassifierNative
      • FitMetadata
      • HUGIMLTuneResult
    • RPTE downstream model
      • leaf_path_conditions()
      • leaf_path_conditions_with_thresholds()
      • simplify_conditions()
      • LeafWiseBoundedLookaheadRPTEFeatureExtractor
      • LeafWiseBoundedLookaheadRPTEFeatureLR
      • aggregate_rule_table_by_raw_source()
      • format_rpte_rule_tree()
      • rpte_rule_tree_sections()
    • Complexity and inspection
      • get_complexity()
      • get_complexity_report()
      • get_instance_inspection_units()
    • Adaptive binning
      • Adaptive-binning algorithm (three steps)
      • Non-finite value handling
      • Usage
      • HUGIMLAdaptive
    • Metrics
      • Quick reference
      • Available metrics
      • InterpretabilityMetrics
      • compute_all_metrics()
      • metrics_dataframe()
    • Hyperparameter configuration
      • make_l1_logistic_base_estimator()
      • get_hugiml_grid()
      • list_hugiml_grids()
      • get_baseline_grid()
      • get_budgeted_baseline_grid()
    • Calibration
      • CalibrationResult
      • evaluate_calibration()
      • reliability_diagram_data()
      • brier_decomposition()
    • Plots
      • Public API
      • HUGPlotter
    • Governance
      • ModelCard
      • AuditArtifact
      • GovernanceMetadata
      • generate_model_card()
      • package_audit_artifacts()
    • Explainability
      • ExplainabilityReport
      • FeatureLineage
      • ExplanationStabilityMetrics
      • HUGPatternExplainer
      • shap_values_from_pattern_matrix()
    • Monitoring
      • PredictionMonitor
      • DriftDetector
      • DriftReport
    • Multiclass and imbalance
      • MulticlassHUGReport
      • make_imbalanced_pipeline()
      • encode_high_cardinality()
      • apply_encoding()
    • Pattern pruning
      • Workflow
      • PatternEditor
      • RemovalRecord
    • Serialization
      • Format (v3+ — current writer)
      • Authentication
      • Backward compatibility (v1/v2)
      • save_model()
      • load_model()
      • generate_sbom()
    • Telemetry
      • OpenTelemetry
      • Prometheus
      • Debug logging
      • HUGIMLTracer
      • HUGIMLMetrics
      • instrument_classifier()
    • Exceptions
      • HUGIMLError
      • HUGIMLFitError
      • HUGIMLMiningError
      • HUGIMLTimeoutError
      • HUGIMLMemoryError
      • HUGIMLValidationError
      • HUGIMLSchemaError
      • HUGIMLParamError
      • HUGIMLSerializationError
      • HUGIMLVersionError
      • HUGIMLPredictionError
      • HUGIMLWarning
      • HUGIMLConvergenceWarning
      • HUGIMLDtypeDriftWarning
      • HUGIMLRangeWarning
      • HUGIMLDegradedWarning
      • HUGIMLDeprecationWarning
    • Dashboard modules
      • SimpleTuneResult
      • PrunedRepresentationResult
      • fit_hugiml_config()
      • fit_feature_pruned_hugiml()
      • fit_representation_pruned_downstream()
      • ModelSpec
      • RuleFitClassifierAdapter
      • dataframe_for_display()
      • adaptive_binning_table()
      • feature_shape_frame()
      • plot_feature_shape()
      • render_feature_effect_profiles()
      • render_adaptive_binning_evidence()
      • survivor_led_patterns_frame()
      • render_survivor_led_pattern_audit()
      • rpte_rule_evidence_frame()
      • render_rpte_rule_evidence()
      • render_augmented_pair_traceability()
      • rpte_is_active()
      • rpte_has_tree_representation()
      • rpte_source_feature_names()
      • rpte_direct_source_terms_frame()
      • rpte_split_usage_frame()
      • rpte_source_inventory_frame()
      • rpte_raw_input_lineage_frame()
      • rpte_representation_flow_frame()
      • rpte_model_comparison_row()
    • LLM assistant add-on
      • DatasetInfo
      • ActionRequest
      • ActionResult
      • DatasetRegistry
      • MemoryInfo
      • ModelProfile
      • ModelOption
      • get_profiles()
      • get_model_catalog()
      • estimated_model_size_b()
      • is_lightweight_supported_model()
      • is_below_minimum_llm_model()
      • get_memory_info()
      • recommend_profile()
      • model_availability()
      • check_ollama()
      • plan_request()
      • ModelSession
      • HUGIMLActionOrchestrator
      • generate_qna_html()
  • Classifier architecture
    • Public contract
    • Internal modules
    • Dependency rules
    • Adding functionality

Project

  • Contributing
  • Changelog
hugiml-core
  • Search

© Copyright 2026, Srikumar Krishnamoorthy.

Built with Sphinx using a theme provided by Read the Docs.