LMKL-Net: A Fast Localized Multiple Kernel Learning Solver via Deep Neural Networks
This work addresses the computational efficiency and accuracy challenges in kernel-based learning for machine learning practitioners, representing an incremental improvement over existing MKL solvers.
The paper tackles the problem of localized multiple kernel learning (LMKL) by proposing LMKL-Net, a deep neural network that parameterizes the gating function and classifier, resulting in improved accuracy and training that is about two orders of magnitude faster with smaller memory footprint on benchmark datasets.
In this paper we propose solving localized multiple kernel learning (LMKL) using LMKL-Net, a feedforward deep neural network. In contrast to previous works, as a learning principle we propose {\em parameterizing} both the gating function for learning kernel combination weights and the multiclass classifier in LMKL using an attentional network (AN) and a multilayer perceptron (MLP), respectively. In this way we can learn the (nonlinear) decision function in LMKL (approximately) by sequential applications of AN and MLP. Empirically on benchmark datasets we demonstrate that overall LMKL-Net can not only outperform the state-of-the-art MKL solvers in terms of accuracy, but also be trained about {\em two orders of magnitude} faster with much smaller memory footprint for large-scale learning.