ECON: a Kernel Basis Pursuit Algorithm with Automatic Feature Parameter Tuning, and its Application to Photometric Solids Approximation
This paper introduces a new algorithm, namely the Equi- Correlation Network (ECON), to perform supervised clas- siﬁcation, and regression. ECON is a kernelized LARS-like algorithm, by which we mean that ECON uses an l1 reg- ularization to produce sparse estimators, ECON efﬁciently rides the regularization path to obtain the estimator associ- ated to any regularization constant values, and ECON rep- resents the data by way of features induced by a feature function. The originality of ECON is that it automatically tunes the parameters of the features while riding the regu- larization path. So, ECON has the unique ability to pro- duce optimally tuned features for each value of the constant of regularization. We illustrate the remarkable experimen- tal performance of ECON on standard benchmark datasets; we also present a novel application of machine learning in the ﬁeld of computer graphics, namely the approximation of photometric solids.