Constructing Orthogonal Latent Features for Arbitrary Loss
نویسندگان
چکیده
A boosting framework for constructing orthogonal features targeted to a given loss function is developed. Combined with techniques from spectral methods such as PCA and PLS, an orthogonal boosting algorithm for linear hypothesis is used to efficiently construct orthogonal latent features selected to optimize the given loss function. The method is generalized to construct orthogonal nonlinear features using the kernel trick. The resulting method, Boosted Latent Features (BLF) is demonstrated to both construct valuable orthogonal features and to be a competitive inference method for a variety of loss functions. For the least squared loss, BLF reduces to the PLS algorithm and preserves all the attractive properties of that algorithm. As in PCA and PLS, the resulting nonlinear features are valuable for visualization, dimensionality reduction, improving generalization by regularization, and use in other learning algorithms, but now these features can be targeted to a specific inference task/loss function. The data matrix is factorized by the extracted features. The low-rank approximation of the data matrix provides efficiency and stability in computation, an attractive characteristic of PLS-type methods. Computational results demonstrate the effectiveness of the approach on a wide range of classification and regression problems.
منابع مشابه
Quasi-Optimal and Optimal Generalized Mutually Orthogonal ZCZ Sequence Sets Based on an Interleaving Technique
The present paper proposes methods for constructing quasi-optimal or optimal generalized mutually orthogonal zerocorrelation zone (GMO-ZCZ) sequence sets. Zero-correlation zone (ZCZ) sequence sets have been studied as spreading sequences for approximately synchronized code-division multiple-access (ASCDMA) systems. A mutually orthogonal ZCZ (MO-ZCZ) sequence set is composed of several small ZCZ...
متن کاملThermal Development for Ducts of Arbitrary Cross Sections by Boundary-Fitted Coordinate Transformation Method
The non-orthogonal boundary-fitted coordinate transformation method is applied to the solution of steady three-dimensional momentum and energy equations in laminar flow to obtain temperature field and Nusselt numbers in the thermal entry region of straight ducts of different cross sectional geometries. The conservation equations originally written in Cartesian coordinates are parabolized in the...
متن کاملSimulation of Styrene Polymerization in Arbitrary Cross-Sectional Duct Reactors by Boundary-Fitted Coordinate Transformation Method
The non-orthogonal boundary-fitted coordinate transformation method is applied to the solution of steady three-dimensional conservation equations of mass, momentum, energy and speciescontinuity to obtain the laminar velocity, temperature and concentration fields for simulation of polymerization of styrene in arbitrary cross-sectional duct reactors. Variable physical properties (except for speci...
متن کاملNumerical Solution of Reacting Laminar Flow Heat and Mass Transfer in Ducts of Arbitrary Cross-Sections for Newtonian and Non-Newtonian Fluids
This study is concerned with the numerical analysis, formulation, programming and computation of steady, 3D conservation equations of reacting laminar flow heat and mass transfer in ducts of arbitrary cross-sections. The non-orthogonal boundary-fitted coordinate transformation method is applied to the Cartesian form of overall-continuity, momenta, energy and species-continuity equations, parabo...
متن کاملAn Algorithm for Constructing All Families of Codes of Arbitrary Requirement in an OCDMA System
—A novel code construction algorithm is presented to find all the possible code families for code reconfiguration in an OCDMA system. The algorithm is developed through searching all the complete subgraphs of a constructed graph. The proposed algorithm is flexible and practical for constructing optical orthogonal codes (OOCs) of arbitrary requirement. Simulation results show that one should cho...
متن کامل