A Simple Bounding Box for Kernelized Log-Linear Regression and its Implications – In this article, we review the performance of a new learning-based method for the classification of binary classification problems. Our method is based on learning Bayes’ generalized log-Linear regression (LLRL) to classify data with a linear class model. In particular, we use a variational inference procedure to derive a Bayes projection from the log-Linear regression. Our method is shown to be effective for classification problems when the linear class model for the data is a linear LER model. Experimental results validate our method for classification problems that do not contain a linear class, such as classification under the presence of a binary class. To the best of our knowledge, this study is the first to test our method using binary data.
This paper addresses the problem of learning a high-dimensional continuous graph from data. Rather than solving the problem of sparse optimization, we propose a novel technique for learning the graph from data. Our approach is based on a variational approach that is independent of the data. This is motivated by the observation that high-dimensional continuous graphs tend to be chaotic and sparse, which has been observed previously. We show that when the graph is not convex, it can also be represented by a finite-dimensional subgraph.
On the Generalizability of Kernelized Linear Regression and its Use as a Modeling Criterion
Optimal Topological Maps of Plant Species
A Simple Bounding Box for Kernelized Log-Linear Regression and its Implications
A Novel Distance-based Sparse Tensor Factorization for Spectral Graph Classification
A Hybrid Learning Framework for Discrete Graphs with Latent VariablesThis paper addresses the problem of learning a high-dimensional continuous graph from data. Rather than solving the problem of sparse optimization, we propose a novel technique for learning the graph from data. Our approach is based on a variational approach that is independent of the data. This is motivated by the observation that high-dimensional continuous graphs tend to be chaotic and sparse, which has been observed previously. We show that when the graph is not convex, it can also be represented by a finite-dimensional subgraph.