L0 Regularization Based Neural Network Design and Compression

05/31/2019
by   S. Asim Ahmed, et al.
0

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.

READ FULL TEXT

Please sign up or login with your details

Forgot password? Click here to reset