voiage.methods.perspective_tensor

Decision-theoretic Value of Perspective reference implementation.

This module treats strategies as the alternatives and perspectives as evaluative lenses. The primary quantity is current-information directional expected perspective regret. Per-draw regret is exposed only as a diagnostic. The module also provides tie-aware perspective acceptability probabilities and an exact expected-value frontier over convex mixtures of two perspectives.

Classes

• PerspectiveError Invalid perspective-analysis input.
• DecisionRule Supported decision rules.
• TiePolicy How exact or numerical ties are handled.
• PerspectiveRegret Directional loss from choosing under one lens and evaluating under another.
• PerspectiveAcceptabilityRow Probability of optimality and expected-value rank for one strategy.
• PerspectiveMixtureFrontierRow Strategy statistics at a sampled mixture weight.
• PerspectiveFrontierSegment An exact interval of perspective weights with the same expected-value optimum.
• PerspectiveSwitchPoint Change of optimal strategies along a two-perspective mixture.
• NetBenefitTensor Dense draw x strategy x perspective net-benefit tensor.

Functions

• expected_net_benefit Average each strategy’s net benefit over draws.
• evop Evaluate directional regret with optional population scaling.
• regret_matrix Return directional regret records for ordered perspective pairs.
• perspective_acceptability_frontier Compute draw-level optimality probabilities and expected-value ranks.
• with_weighted_perspective Append a weighted combination of existing perspectives.
• perspective_mixture_frontier Evaluate strategy performance on a regular two-perspective weight grid.
• exact_perspective_frontier Partition the mixture interval by optimal expected net benefit.
• expected_value_switch_points Compatibility helper.
• mcda_feature_records Collect benefits, ranks, discordance and regret by lens and strategy.
• tensor_from_records Build a dense labelled tensor from long-form net benefit records.