Replay program for depth reduction steps with its G draws read from trace
instead of sampled, still integrating every E draw: the law of the real the
replay returns exactly at depth. The result is the zero measure when the trace does not fit
the program: a G site that finds no entry or an entry of another primitive, or an
entry left over when the program returns. For discrete, the comparison op = op'
checks the number of supplied probabilities; their values are evaluated operands.
Equations
- One or more equations did not get rendered due to their size.
- Determinize.Proof.Traces.outputGivenTraceAt 0 (Determinize.Spec.Paper.Expr.real ℝ Determinize.Spec.Paper.DistributionAction value) [] = MeasureTheory.Measure.dirac value
- Determinize.Proof.Traces.outputGivenTraceAt 0 x✝¹ x✝ = 0
Instances For
The law of the output of program given that its G draws were trace: the
program replayed along the trace, at whichever depth it returns.
Equations
- Determinize.Proof.Traces.outputGivenTrace program trace = MeasureTheory.Measure.sum fun (depth : ℕ) => Determinize.Proof.Traces.outputGivenTraceAt depth program trace
Instances For
Draw a trace from traces, then an output from outputs trace: the joint law of the pair.
Equations
- One or more equations did not get rendered due to their size.
Instances For
The mean of the source replay law at a fixed generation trace.
Equations
- Determinize.Proof.Traces.replayMean program trace = ∫ (value : ℝ), value ∂Determinize.Proof.Traces.outputGivenTrace program trace