Forward-Backward Selection with Early Dropping
نویسندگان
چکیده
Forward-backward selection is one of the most basic and commonly-used feature selection algorithms available. It is also general and conceptually applicable to many different types of data. In this paper, we propose a heuristic that significantly improves its running time, while preserving predictive accuracy. The idea is to temporarily discard the variables that are conditionally independent with the outcome given the selected variable set. Depending on how those variables are reconsidered and reintroduced, this heuristic gives rise to a family of algorithms with increasingly stronger theoretical guarantees. In distributions that can be faithfully represented by Bayesian networks or maximal ancestral graphs, members of this algorithmic family are able to correctly identify the Markov blanket in the sample limit. In experiments we show that the proposed heuristic increases computational efficiency by about two orders of magnitude in high-dimensional problems, while selecting fewer variables and retaining predictive performance. Furthermore, we show that the proposed algorithm and feature selection with LASSO perform similarly when restricted to select the same number of variables, making the proposed algorithm an attractive alternative for problems where no (efficient) algorithm for LASSO exists.
منابع مشابه
ILU and IUL factorizations obtained from forward and backward factored approximate inverse algorithms
In this paper, an efficient dropping criterion has been used to compute the IUL factorization obtained from Backward Factored APproximate INVerse (BFAPINV) and ILU factorization obtained from Forward Factored APproximate INVerse (FFAPINV) algorithms. We use different drop tolerance parameters to compute the preconditioners. To study the effect of such a dropping on the quality of the ILU ...
متن کاملComparative Approach to the Backward Elimination and for-ward Selection Methods in Modeling the Systematic Risk Based on the ARFIMA-FIGARCH Model
The present study aims to model systematic risk using financial and accounting variables. Accordingly, the data for 174 companies in Tehran Stock Exchange are extracted for the period of 2006 to 2016. First, the systematic risk index is estimated using the ARFIMA-FIGARCH model. Then, based on the research background, 35 affective financial and accounting variables are simultaneously used with t...
متن کاملModel Selection in Linear Mixed Effects Models Using SAS PROC MIXED
Although there are disadvantages associated with model building procedures such as backward, forward and stepwise procedures (e.g. multiple testing, arbitrary significance level used in dropping or acquiring variables), many analysts use these procedures and are not aware that alternative modeling selection methods exist. This paper focuses on model selection using the Akaike Information Criter...
متن کاملInterleaving Forward Backward Feature Selection
Selecting appropriate features has become a key task when dealing with high-dimensional data. We present a new algorithm designed to find an optimal solution for classification tasks. Our approach combines forward selection, backward elimination and exhaustive search. We demonstrate its capabilities and limits using artificial and real world data sets. Regarding artificial data sets interleavin...
متن کاملWritten word recognition by the elementary and advanced level Persian-English bilinguals
According to a basic prediction made by the Revised Hierarchical Model (RHM), at early stages of language acquisition, strong L2-L1 lexical links are formed. RHM predicts that these links weaken with increasing proficiency, although they do not disappear even at higher levels of language development. To test this prediction, two groups of highly proficie...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید
ثبت ناماگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید
ورودعنوان ژورنال:
- CoRR
دوره abs/1705.10770 شماره
صفحات -
تاریخ انتشار 2017