Skip to content

Cross-Pillar Synergy Engine (CONCEPT:KG-2.4)

Overview

Discovers non-obvious functional synergies between the 5 Unified Pillars by analyzing concept bridges, computing pillar coupling metrics, and suggesting missing relationships. Leverages the Analogy Engine (KG-2.7), SKOS taxonomy, and transitive OWL properties. OWL property: hasSynergyWith (symmetric, defined in ontology_quant.ttl).

Implementation Details

  • Source Code: agent_utilities/knowledge_graph/core/synergy_engine.py
  • Pillar: KG

Documentation Coverage

This is an auto-generated dedicated concept page to ensure 100% documentation coverage across the ecosystem.

Formal Relations Engine (CONCEPT:KG-2.6)

Overview

Mathematical relation properties (Reflexive, Symmetric, Transitive closures) and Equivalence Classes from MCS Ch 4. Provides zero-shot entity resolution by formally defining equivalence sets.

Implementation Details

  • Source Code: agent_utilities/knowledge_graph/core/inference_engine.py (closure rules), agent_utilities/knowledge_graph/core/owl_bridge.py (symmetric/transitive_closure inference)
  • Pillar: KG

Documentation Coverage

This is an auto-generated dedicated concept page to ensure 100% documentation coverage across the ecosystem.

State Machine Invariant Engine (CONCEPT:KG-2.6)

Overview

Deterministic Finite Automata (DFA) abstractions and provable state invariants from MCS Ch 6. Formally validates transitions against structural invariants, preventing infinite loops. Implemented as the FormalStateMachine class.

Implementation Details

  • Source Code: agent_utilities/knowledge_graph/core/formal_reasoning_core.py
  • Pillar: KG

Documentation Coverage

This is an auto-generated dedicated concept page to ensure 100% documentation coverage across the ecosystem.

Markov Transition Forecasting (CONCEPT:KG-2.6)

Overview

Markov Chain transition matrices over agent interaction traces (Vectorized Topologies) from MCS Ch 21. Calculates stationary distribution (Eigenvector) to predict statistical failure nodes.

Implementation Details

  • Source Code: agent_utilities/knowledge_graph/core/formal_reasoning_core.py
  • Pillar: KG

Core Capabilities

The MarkovTransitionModel provides: - Trace ingestion — Build transition matrices from sequential state observations - Stationary distribution — Power iteration to find long-run state probabilities - Sink node prediction — Identify absorbing/terminal states - Top-k next-state prediction — Predict most likely successor states (used by PreemptiveCacheEngine) - Chapman-Kolmogorov n-step forecasting — Multi-step transition probabilities via matrix powers - State-specific forecasting — Probability distribution over states after n steps from a given starting state

Documentation Coverage

This is an auto-generated dedicated concept page to ensure 100% documentation coverage across the ecosystem.

Markov Regime Detection (CONCEPT:KG-2.6)

Overview

Market regime detection and forecasting using Markov Chains for financial time-series analysis. Classifies market states (Bull/Bear/Sideways) from returns data and builds probabilistic transition models for regime forecasting, trading signal generation, and walk-forward backtesting.

Architecture

Raw Returns → MarketRegimeDetector → State Labels
                                          │
                                          ▼
              MarkovRegimeModel ← MarkovTransitionModel
                    │                     │
                    ▼                     ▼
             Regime Forecast        KG Persistence
                    │          (FinanceEngineMixin)
                    ▼
            Trading Signal / Walk-Forward Backtest

Implementation Details

  • Source Code: agent_utilities/knowledge_graph/core/markov_regime.py
  • Engine Integration: agent_utilities/knowledge_graph/orchestration/engine_finance.py
  • Domain Models: agent_utilities/models/domains/finance.py
  • Pillar: KG + Finance Domain

Key Classes

MarketRegimeDetector

Detects market regimes from returns time-series. Supports two rolling return methods: - rolling_sum: Simple sum of daily returns (fast, good approximation for small returns) - compounding: Proper ∏(1+r_i) - 1 (more accurate for volatile assets like crypto)

MarkovRegimeModel

End-to-end Markov Chain model: - fit(returns) — Detect regimes → build transition matrix - forecast(state, n_steps) — N-step regime probabilities - generate_signal(state) — Bull_prob - Bear_prob trading signal - walk_forward_backtest(returns, lookback) — Rolling re-estimation with no lookahead bias - to_kg_properties() — Serialize model state for KG persistence

HiddenMarkovRegimeModel

Gaussian Hidden Markov Model for latent regime detection (requires hmmlearn): - fit(returns) — Baum-Welch EM with multiple random restarts - decode(returns) — Viterbi decoding of most likely regime sequence - predict_proba(returns) — Posterior regime probabilities per timestep

Asset-Class-Specific Defaults

Asset Class Bull Threshold Bear Threshold Window
Equities +2.0% -2.0% 20
Crypto +5.0% -5.0% 14
Forex +0.5% -0.5% 30
Commodities +1.5% -1.5% 25
Fixed Income +0.3% -0.3% 40

All thresholds are configurable at instantiation.

KG Node Types

  • MarkovRegimeStateNode — A detected regime observation
  • MarkovTransitionMatrixNode — A serialized transition matrix
  • RegimeSignalNode — A generated trading signal

KG Edge Types

  • TradingStrategy -[:MODELS_REGIME]-> MarkovTransitionMatrix
  • MarkovTransitionMatrix -[:DETECTED_REGIME]-> MarkovRegimeState
  • TradingStrategy -[:GENERATES_SIGNAL]-> RegimeSignal

Service Registry

Registered as discoverable services in the ServiceRegistry (CONCEPT:ORCH-1.4): - markov_regime_detection (Layer: domain, Domain: finance) - hmm_regime_detection (Layer: domain, Domain: finance)