HadaNets: Flexible Quantization Strategies for Neural Networks
Abstract
On-board processing elements on UAVs are currently inadequate for training and inference of Deep Neural Networks. This is largely due to the energy consumption of memory accesses in such a network. HadaNets introduce a flexible train-from-scratch tensor quantization scheme by pairing a full precision tensor to a binary tensor in the form of a Hadamard product. Unlike wider reduced precision neural network models, we preserve the train-time parameter count, thus out-performing XNOR-Nets without a train-time memory penalty. Such training routines could see great utility in semi-supervised online learning tasks. Our method also offers advantages in model compression, as we reduce the model size of ResNet-18 by 7.43 times with respect to a full precision model without utilizing any other compression techniques. We also demonstrate a 'Hadamard Binary Matrix Multiply' kernel, which delivers a 10-fold increase in performance over full precision matrix multiplication with a similarly optimized kernel.
Turn this paper into a lesson
ArcXiv compiles a structured reading guide from this paper's metadata: plain-English importance, contributions, prerequisite concepts, which sections to read first, flashcards, and a quiz. Grounded in the abstract, never invented.