|
|
|
|
|
|
|
|
|
Overfitting
| Глоссарий по искусственному интеллекту |
- A term used in neural networks, recursive partitioning, and other automated modeling areas. if the training data is considered to consist both signal and noise (i.e., noisy data), a modeling technique has begun to overfit when it begins capturing the "noise" instead of the "signal." this usually occurs when a model is allowed to increase its number of "parameters`` such as regression coefficients, splits in recursive partitioning schemes, or hidden units in neural networks. the effect of overfitting is to reduce the applicability of the model to other data sets, (i.e., to limit its generalizability). in the extreme, the model can only "predict" its input. the most straight-forward way to eliminate this effect is to require a very large number of observations per parameter in the model.
- A problem in data mining when random variations in data are misclassified as important patterns. overfitting often occurs when the data set is too small to represent the real world.
|
Переобучение, русский
Ситуация, когда на обучающей последовательности ошибки сети были очень малы, но на новых данных становятся большими.
|
Partitioning, английский
- Расчленение
- Разделение; создание разделов; разбиение на разделы; выделение разделов; организация разделов; декомпозиция; секционирование
- The process of replacing a table with multiple smaller tables.
Parameters, английский
Параметры
Regression, английский
- Statistical technique used to evaluate relationships among variables (22).
- Регрессия
- 1. a stage where symptoms of a disease are disappearing and the person is getting better 2. (in psychiatry) the process of returning to a mental state which existed when the person was younger
- Regression commonly refers to the process of developing an empirical (data-driven) model to predict and/or explain one or more attributes in a database or set of data. it is most frequently associated with the simple linear model (y=mx+b) taught in most introductory statistics courses; the same ideas have been extended in many directions, including classification problems. when the emphasis is on hypothesis testing and simple models, the regression output is typically a few parameters that provide a direct linkage from the input variables to the predicted variables (or classification). in other situations the emphasis is on explaining as much of the variability in the output variables as is "reasonable" from the input variables. in this case, there are a number of "advanced" techniques, such as smoothing splines, decision trees, neural nets, and so forth, for which there are many "free" parameters. the meaning of any one of these parameters can be obscure. many data mining techniques are, at their core, variations on well-known regression techniques. see also: classification, clustering, decision trees, neural nets.
- The reappearance of a previously fixed problem.
- The statistical process of predicting one or more continuous variables, such as profit or loss, based on other attributes in the dataset.
- A mathematical technique used to explain and/or predict. the general form is y = a + bx + u, where y is the variable that we are trying to predict; x is the variable that we are using to predict y, a is the intercept; b is the slope, and u is the regression residual. the a and b are chosen in a way to minimize the squared sum of the residuals. the ability to fit or explain is measured by the r-square.
- A seaward retreat of a shoreline, generally expressed as a seaward
Applicability, английский
- Применимость
- Применимость; пригодность
- Пригодность, применимость
Generalizability, английский
Observations, английский
Замечания эксперта (по существу заявки)
|
Local operators, английский
A term used in feature analysis to refer to functions that operate in a restricted neighborhood around a given item (e.g., a pixel in a picture).
Signal state space, английский
A term used in the discussion of neural networks, a signal state space is the space defined by all possible outputs of model. if the model had one continuous bounded output, its state space would be that bounded interval. with two such outputs, the state space can form a rectangle with three a hyper-rectangle or cube, and so on.
|
|
|
|
|
|
|
