|
|
|
|
|
|
|
|
|
Optimal factoring problem
| Глоссарий по искусственному интеллекту |
Probabilistic networks offer a very flexible way to represent uncertainty and its propagation through attributes. however, it can be very "expensive" to compute marginal and joint probabilities from large and complicated nets. one solution is to find an optimal factoring of the network for a given set of target nodes in the network that minimizes the cost of computing the target probabilities. see also: bayesian network.
|
|
Probabilistic, английский
- Вероятностный
- Attribute of non-deterministic systems whose transitions between states follow known or ascertainable probabilities (->probability). the ergodic behavior of probabilistic systems is describable as a markov chain. probabii.ity
Uncertainty, английский
- Неизвестность. неопределенность.
- Неопределенность. недостоверность. сомнительность. ненадежность.
- Неопредел
- Неопределенность. параметр, связанный с результатом измерения, характеризующий дисперсию значений, которые могут быть обоснованно отнесены к аналиту. таким параметром может быть, например, среднеквадратическое откло-нение (или кратное ему) или ширина доверительного интервала. также: оценка, относящаяся к результату проверки, которая характеризует область значений, внутри которой лежит истинное значение.
- The (average) number of binary decisions a decision maker has to make in order to select one out of a set of mutually exclusive alternatives, a measure of an observer`s ignorance or lack of information (->bit). since the categories within which events are observed are always specified by an observer, the notion of uncertainty emphasises the cognitive dimension of information processes, specifically in the form of measures of variety, statistical entropy including noise and equivocation. 78
- Неопределенность. параметр, связанный с результатом измерения, характеризующий дисперсию значений, которые могут быть обоснованно отнесены к аналиту. таким параметром может быть, например, среднеквадратическое отклонение (или кратное ему) или ширина доверительного интервала. также: оценка, относящаяся к результату проверки, которая характеризует область значений, внутри которой лежит истинное значение.
Propagation, английский
- Прохождение, распространение (напр, радиоволн)
- An act of causing something to spread or multiply
- Распространение; размножение; продвижение; передача
- Распространение
- The process of distributing an index from a content index server to one or more web servers for the purposes of providing search.
- Advancement of a wave through a medium.
- Advancement of energy or a crack through a medium. see also brittle crack propagation; ductile crack propagation; fatigue crack propagation.
- Movement of a wave through a medium.7,21
Complicated, английский
|
Probabilistic inference, английский
Probabilistic inference is the process of inferring the probabilities of interest in a (bayesian) model conditional on the values (configuration) of other attributes in the model. for simpler models, such as a regression tree or decision tree model, this can be a fairly straightforward problem. for more complex models, such as a bayesian network, exact inference can be quite difficult, and may require either approximations or sampling based approaches, such as markov chain monte carlo (mcmc) methods. see also: gibbs sampling, simulated annealing.
Helmholtz machine, английский
Probability propagation on a large, multiply-connected bayesian network can be computationally difficult. a helmholtz machine tackles this problem by coupling the original generative network with a second recognition network that produces a quick approximation to the desired solution. different sets of visible variables in the network can require different recognizer networks. recognizer networks can be divided into factorial networks which assume a simple naive bayes model, given the visible variables, and nonfactorial models which allow for a more complicated relationship among the hidden variables given the visible variables. see also: generalized em, markov chain monte carlo methods, sum-product algorithm, variational inference.
|
|
|
|
|
|
|
