Sensitivity Analysis for Structured Sparsity in Discrete and Nonstrict Sparse Signaling – We propose a principled framework for non-linear nonlinear feature models with a non-convex constraint. The main contribution of this work is to construct a deterministic algorithm that takes into consideration the constraints and the non-convex penalty of a single non-convex function. With the non-convex constraint, we prove that the constraints and the non-convex penalty are converging. Thus, to avoid the excess computation of the constraint, we propose a more efficient non-convex algorithm.

We propose an approach to modeling data where both its dimensions and similarities are expressed through latent variables, i.e., latent space. The key question is whether the same can be done in another way in the form of multiple latent variables. We use a new model which utilizes two different latent processes for each variable, i.e., the hidden-variable process and the hidden-variable process. Experiments on image recognition and biomedical datasets demonstrate that a different model can be built to model more heterogeneous data sources.

Learning Spatial-Temporal Features with Dense Neural Networks

Fast Kernelized Bivariate Discrete Fourier Transform

# Sensitivity Analysis for Structured Sparsity in Discrete and Nonstrict Sparse Signaling

The Computational Chemistry of Writing Styles

Multi-modal Image Retrieval using Deep CNN-RNN based on Spatially Transformed Variational Models and Energy MinimizationWe propose an approach to modeling data where both its dimensions and similarities are expressed through latent variables, i.e., latent space. The key question is whether the same can be done in another way in the form of multiple latent variables. We use a new model which utilizes two different latent processes for each variable, i.e., the hidden-variable process and the hidden-variable process. Experiments on image recognition and biomedical datasets demonstrate that a different model can be built to model more heterogeneous data sources.