CVLGDec 15, 2017

Reducing Deep Network Complexity via Sparse Hierarchical Fourier Interaction Networks

arXiv:1801.01451v35 citations
Originality Synthesis-oriented
AI Analysis

This addresses complexity reduction in deep networks for visual and linguistic tasks, but appears incremental as it builds on existing frequency domain methods.

The paper tackles the problem of deep network complexity by introducing Sparse Hierarchical Fourier Interaction Networks, which reduce complexity through a hierarchical Fourier transform and sparse masking, achieving unspecified compression gains.

This paper presents a Sparse Hierarchical Fourier Interaction Networks, an architectural building block that unifies three complementary principles of frequency domain modeling: A hierarchical patch wise Fourier transform that affords simultaneous access to local detail and global context; A learnable, differentiable top K masking mechanism which retains only the most informative spectral coefficients, thereby exploiting the natural compressibility of visual and linguistic signals.

Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes