Robust Networks: Neural Networks Robust to Quantization Noise and Analog Computation Noise Based on Natural Gradient
Abstract | Deep neural networks (DNNs) have had tremendous success in a variety of statistical learning applications due to their vast expressive power. Most applications run DNNs on the cloud on parallelized architectures. There is a need for for efficient DNN inference on edge with low precision hardware and analog accelerators. To make trained models more robust for this setting, quantization and analog compute noise are modeled as weight space perturbations to DNNs and an information theoretic regularization scheme is used to penalize the KL-divergence between perturbed and unperturbed models. This regularizer has similarities to both natural gradient descent and knowledge distillation, but has the advantage of explicitly promoting the ne... (more) |
---|---|
Created Date | 2019 |
Contributor | Kadambi, Pradyumna (Author) / Berisha, Visar (Advisor) / Dasarathy, Gautam (Committee member) / Seo, Jae-Sun (Committee member) / Cao, Yu (Committee member) / Arizona State University (Publisher) |
Subject | Artificial intelligence / Computer engineering / Computer science / analog neural network / distillation / fisher information / kl-divergence / non-volatile memory accelerator / quantized neural network |
Type | Masters Thesis |
Extent | 83 pages |
Language | English |
Copyright |
|
Note | Masters Thesis Computer Engineering 2019 |
Collaborating Institutions | Graduate College / ASU Library |
Additional Formats | MODS / OAI Dublin Core / RIS |