LGMLMay 31, 2019

L0 Regularization Based Neural Network Design and Compression

arXiv:1905.13652v1
Originality Synthesis-oriented
AI Analysis

This work addresses complexity reduction for scaling AI to real-world, off-the-cloud applications, but it is incremental as it applies an existing regularization method to new contexts.

The paper tackled the problem of massive over-parameterization in deep neural networks, which leads to issues like adversarial susceptibility and high costs, by applying L0 regularization to reduce complexity and analyze trade-offs, showing it captures input saliency on MNIST and signal modulation datasets.

We consider complexity of Deep Neural Networks (DNNs) and their associated massive over-parameterization. Such over-parametrization may entail susceptibility to adversarial attacks, loss of interpretability and adverse Size, Weight and Power - Cost (SWaP-C) considerations. We ask if there are methodical ways (regularization) to reduce complexity and how can we interpret trade-off between desired metric and complexity of DNN. Reducing complexity is directly applicable to scaling of AI applications to real world problems (especially for off-the-cloud applications). We show that presence and evaluation of the knee of the tradeoff curve. We apply a form of L0 regularization to MNIST data and signal modulation classifications. We show that such regularization captures saliency in the input space as well.

Foundations

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

Your Notes