Abstract: The iterative hard thresholding (IHT) algorithm is widely used for recovering sparse signals in compressed sensing. Despite the development of numerous variants of this effective algorithm, ...

We propose a novel neural attention architecture to tackle machine comprehension tasks, such as answering Cloze-style queries with respect to a document. Unlike previous models, we do not collapse the ...

Conditional text-to-image generation approaches commonly focus on generating a single image in a single step. One practical extension beyond one-step generation is an interactive system that generates ...

Abstract: Traditional superimposed single pilot (SSIP) and superimposed sparse pilot (SSPP) schemes require high-power pilot symbols for channel estimation, leading to a high peak-to-average power ...