Optimal Probability Estimation with Applications to Prediction and Classification
نویسندگان
چکیده
Via a unified view of probability estimation, classification, and prediction, we derive a uniformlyoptimal combined-probability estimator, construct a classifier that uniformly approaches the error of the best possible label-invariant classifier, and improve existing results on pattern prediction and compression. ∗[email protected] †[email protected] ‡[email protected] §[email protected]
منابع مشابه
Compression-based methods for nonparametric on-line prediction, regression, classification and density estimation of time series
Jorma Rissanen has discovered some deep connections between universal coding (or universal data compression) and mathematical statistics. In particular, the MDL principle has been one of the most powerful methods of modern mathematical statistics. In this paper we apply Rissanen’s approach and ideas to some statistical problems concerned with time series. We address the problem of nonparametric...
متن کاملPrediction the Return Fluctuations with Artificial Neural Networks' Approach
Time changes of return, inefficiency studies performed and presence of effective factors on share return rate are caused development modern and intelligent methods in estimation and evaluation of share return in stock companies. Aim of this research is prediction of return using financial variables with artificial neural network approach. Therefore, the statistical population of this study incl...
متن کاملNew Optimal Observer Design Based on State Prediction for a Class of Non-linear Systems Through Approximation
This paper deals with the optimal state observer of non-linear systems based on a new strategy. Despite the development of state prediction in linear systems, state prediction for non-linear systems is still challenging. In this paper, to obtain a future estimation of the system states, initially Taylor series expansion of states in their receding horizons was achieved to any specified order an...
متن کاملPrediction of potential habitat distribution of Artemisia sieberi Besser using data-driven methods in Poshtkouh rangelands of Yazd province
The present study aimed to model potential habitat distribution of A. sieberi, and its ecological requirements using generalized additive model (GAM) and classification and regression tree (CART) in in the Poshtkouh rangelands of Yazd province. For this purpose, pure habitats of the species was delineated and the species presence data was recorded by the systematic-randomize sampling method. Us...
متن کاملN ov 2 00 7 Compression - based methods for nonparametric density estimation , on - line prediction , regression and classification for time series
We address the problem of nonparametric estimation of characteristics for stationary and ergodic time series. We consider finite-alphabet time series and real-valued ones and the following four problems: i) estimation of the (limiting) probability P (u0 . . . us) for every s and each sequence u0 · · · us of letters from the process alphabet (or estimation of the density p(x0, . . . , xs) for re...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
عنوان ژورنال:
دوره شماره
صفحات -
تاریخ انتشار 2013