|
|
|
|
|
|
|
|
|
Результаты поиска термина [likelihood]
|
likelihood | - Вероятность. правдоподобие.
- Правдоподобие; вероятность
- In probabilistic and statistical methods, the likelihood is a measure of the evidence for the data when a hypothesis is assumed to be true. it is commonly used in bayesian and quasi-bayesian techniques. a related measure, minimum message length, can be derived from information theoretic principles. as an example, suppose you observe a data value x that is assumed to be gaussian. the likelihood for that data when the assumed mean is, say, 5, and the variance is 10, is proportional to e-((x-5)2/2*10), the kernel of a gaussian distribution. a likelihood is typically calculated using the kernel, rather than the complete distribution function, which includes the normalizing constants. see also: minimum message length.
| |
likelihood estimation | Оценка степени вероятности | |
likelihood function | Функция правдоподобия | |
likelihood of confusion | - Вероятность спутывания; вероятность смешения; возможность спутывания; возможность смешения (субъективный критерий правомочности товарного знака при наличии сходных товарных знаков на родственные товары)
- Grounds to refuse to register the trademark in your application when it conflicts with a similar trademark that`s registered for goods or services related to yours. if both trademarks were in use in commerce at the same time, it would likely cause confusion for consumers, who may not be able to determine the source of the goods or services.
| |
likelihood of mistake | Вероятность ошибки | |
likelihood ratio | - Отношение правдоподобия
- When evaluating the likelihood of the evidence while comparing two disjoint hypotheses, ha and hb, the likelihood ratio is the ratio of p(e|ha) to p(e|hb). it is sometimes referred to as a bayes factor for ha versus hb the posterior odds for any prior can be computed by multiplying this ratio by the prior odds ratio. see also: odds ratio. linear discriminant analysis (lda) linear discriminant analysis (lda) is a form of supervised learning, originally developed as a statistical tool, for finding an optimal linear rule to separate two classes (concept). the algorithm finds the line in a hyperspace that connects the centroids of the positive and negative examples and chooses a breakpoint along that line that best separates the class, weighting the coefficients to account for any linear correlations between the attributes. when the concept is linearly separable, lda can completely separate the classes. the "linear" in this term refers to the linearity of the function that combines the attributes, not to the attributes themselves. the attribute list can include powers and products of other attributes, as well as other functions such as spline functions of the "real" attribute. the rule is of the form a + b.x + c.y + d.z + ..., i.e., it is linear in the given attributes. the value of a, sometimes called the bias or the intercept, is often chosen so that any positive score on an attribute is taken to mean membership in one class, and a negative score to mean membership in the other. it can be extended to learn multiple classes through a technique called multiple discriminant analysis (mda), which estimates k-1 linear functions to separate k classes on the basis of the attributes. see also: linearly separable, machine learning.
| |
likelihood ratio test |
|
|
|
|
|
|
|
