What an inference result vouches for: its :inference-contract, built
here and nowhere else, from a closed vocabulary. A contract states what
the algorithm guarantees by construction; it is not a check that a run
converged.
:semantics what the particles are (semantics)
:normalizer {:status s} how measure/log-marginal may be read
(normalizer-statuses), with the :estimator it
names when s is :estimate
:weight-contract obligations the weights rest on (weight-contracts)
:adaptation what the algorithm tuned from its own population
(adaptations)
:approximation what it approximated on purpose (approximations)
Construct contracts with weighted-population and markov-chain; read
them with normalizer-status and unbiased-evidence?. An unknown key or
value is refused, so a new guarantee is named here first.
What an inference result vouches for: its `:inference-contract`, built
here and nowhere else, from a closed vocabulary. A contract states what
the algorithm guarantees by construction; it is not a check that a run
converged.
:semantics what the particles are (`semantics`)
:normalizer {:status s} how `measure/log-marginal` may be read
(`normalizer-statuses`), with the `:estimator` it
names when s is :estimate
:weight-contract obligations the weights rest on (`weight-contracts`)
:adaptation what the algorithm tuned from its own population
(`adaptations`)
:approximation what it approximated on purpose (`approximations`)
Construct contracts with `weighted-population` and `markov-chain`; read
them with `normalizer-status` and `unbiased-evidence?`. An unknown key or
value is refused, so a new guarantee is named here first.(adapted contract adaptation)contract noting that the algorithm adapted adaptation from its own
population.
`contract` noting that the algorithm adapted `adaptation` from its own population.
(check contract)contract if every key and value is in the vocabulary, else an error
naming the first that is not.
`contract` if every key and value is in the vocabulary, else an error naming the first that is not.
(markov-chain)(markov-chain {:keys [estimator]})Chain draws' contract: no evidence, unless the chain method computes a
separate estimate by estimator (parallel tempering's stepping stones).
Chain draws' contract: no evidence, unless the chain method computes a separate estimate by `estimator` (parallel tempering's stepping stones).
(normalizer-status result)The normalizer status of a result (anything carrying :inference-contract).
The normalizer status of a result (anything carrying `:inference-contract`).
(unbiased-evidence? result)Whether result's log-marginal is the evidence or an unbiased estimate
of it (on the natural scale): what a pseudo-marginal method may consume.
Whether `result`'s `log-marginal` is the evidence or an unbiased estimate of it (on the natural scale): what a pseudo-marginal method may consume.
(weighted-population)(weighted-population {:keys [semantics normalizer estimator weight-contract
adaptation approximation]
:or {semantics :weighted-population
normalizer :unverified}})A population result's contract: :semantics :weighted-population unless
given, the normalizer :unverified unless :normalizer names another
status (with :estimator for :estimate), and the optional terms given.
A population result's contract: `:semantics` :weighted-population unless given, the normalizer `:unverified` unless `:normalizer` names another status (with `:estimator` for :estimate), and the optional terms given.
cljdoc builds & hosts documentation for Clojure/Script libraries
| Ctrl+k | Jump to recent docs |
| ← | Move to previous article |
| → | Move to next article |
| Ctrl+/ | Jump to the search field |