Open Access
Automatic Parameter Tying in Neural Networks
Reads0
Chats0
TLDR
This paper proposes a novel algorithm that jointly learns and compresses a neural network, and shows that the approach is easy to implement using existing neural network libraries, generalizes L1 and L2 regularization and elegantly enforces parameter tying as well as pruning constraints.Abstract:
Recently, there has been growing interest in methods that perform neural network compression, namely techniques that attempt to substantially reduce the size of a neural network without significant reduction in performance. However, most existing methods are post-processing approaches in that they take a learned neural network as input and output a compressed network by either forcing several parameters to take the same value (parameter tying via quantization) or pruning irrelevant edges (pruning) or both. In this paper, we propose a novel algorithm that jointly learns and compresses a neural network. The key idea in our approach is to change the optimization criteria by adding $k$ independent Gaussian priors over the parameters and a sparsity penalty. We show that our approach is easy to implement using existing neural network libraries, generalizes L1 and L2 regularization and elegantly enforces parameter tying as well as pruning constraints. Experimentally, we demonstrate that our new algorithm yields state-of-the-art compression on several standard benchmarks with minimal loss in accuracy while requiring little to no hyperparameter tuning as compared with related, competing approaches.read more
Citations
More filters
Patent
Voice control of a media playback system
Jonathan P. Lang,Mark Plagge,Simon Jarvis,Romi Kadri,Yean-Nian Willy Chen,Paul Bates,Luis R. Vega-Zayas,Christopher Butts,Nicholas A. J. Millington,Keith Corbin +9 more
TL;DR: In this paper, various aspects of systems and methods for voice control and related features and functionality for various embodiments of media playback devices, networked microphone devices, microphone-equipped media playback device, and speaker-equipped networked microphones devices are disclosed and described.
Patent
Arbitration-based voice recognition
Steven Beckhardt,Ted Lin +1 more
TL;DR: In this paper, a system comprises a first network microphone device (NMD) communicatively coupled to a second NMD via a network interface of the first NMD, wherein the first nMD is configured to: receive, from the second nMD via the network interface, an arbitration message comprising (i) a first measure of confidence associated with a voice input detected by the second NMM and (ii) an identifier associated with at least a portion of the voice inputs detected by a second nMM, determine that the first measures of confidence is greater than a second measure of
Patent
Three-dimensional beam forming with a microphone array
TL;DR: In this paper, a set of received-sound beams are generated by applying a plurality of sets of beamforming coefficients stored in a tangible memory of a network device to sound received via a microphone array of the network device.
Patent
Dynamic computation of system response volume
TL;DR: In this article, the authors determine that a requirement exists to output a system response, and then set a response volume for the first speaker associated with the first networked device and output the system response at the set response volume via the speaker.
Patent
Media playback system with voice assistance
Dayn Wilberding,John Tolomei +1 more
TL;DR: In this article, the authors present a set of techniques for invoking voice assistance for a media playback system, where the system is configured to capture a voice input via at least one microphone device, detect inclusion of one or more of the commands within the voice input, determine that the ones or more commands meets corresponding command criteria associated with the one or multiple commands, and in response to the determination, select a first voice assistant service (VAS) and forego selection of a second VAS.