felipecustodio Overview & Purpose
Decision Networks extend Bayesian networks by incorporating decision nodes and utility nodes. They help agents compute expected utilities of decisions under uncertain environment states.
When to Use
- Decisions under Uncertainty: When actions have uncertain outcomes with distinct financial/operational utilities.
- Risk vs Reward Optimization: Balancing high-risk/high-reward paths against safe/low-reward paths.
Execution Workflow
- Define Chance Nodes: Uncertain state variables in the environment.
- Define Decision Nodes: Specific choices under agent control.
- Define Utility Nodes: Quantifiable payoff or loss functions.
- Compute Expected Utility (EU): $\text{EU}(D) = \sum P(S|D) \times U(S, D)$.
- Select Maximum Utility Action: Choose decision $D^*$ maximizing $\text{EU}(D)$.
Expected Output Contract
### Decision Network Analysis
- **Decision Nodes**: [Available Choices]
- **Chance Nodes**: [Uncertain State Probabilities]
- **Expected Utility**:
- `Choice A`: Expected Utility = X
- `Choice B`: Expected Utility = Y
- **Optimal Choice**: [Choice with Max EU]
Scripts
scripts/decision_networks.py- Deterministic evaluation, state validation, and CLI tool for decision-networks.