Boyu Zhang

Boyu Zhang

Auto-generated profile, based on publicly available information

Boyu Zhang

University of Wisconsin–Madison 1415 Engineering Drive, Madison, WI, USA 53706

Research Focus & Metrics

Academic Career Length

10Years (2017-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
4 May 2018 in IEEE Journal on Emerging and Selected Topics in Circuits and Systems
Verified
Journal Article
Closed access
BZBoyu Zhang
Boyu Zhang
ADAzadeh Davoodi
Azadeh Davoodi
YHYu Hen Hu
Yu Hen Hu
This paper presents a structure simplification procedure that allows efficient energy and accuracy tradeoffs in implementation of trained deep neural networks (DNNs). This structure simplification procedure identifies and eliminates redundant neurons in any layer of the DNN based on the output values of these neurons over the training samples. This procedure may be applied to all layers of a DNN. Pareto-optimal points in terms of accuracy and energy consumption can be easily identified from the configurations across different layers. After redundant neurons are discarded, the weights connect to the remaining neurons will be updated using matrix multiplication without retraining. Yet, retraining may still be applied if desired to further fine tune the performance. In our experiments, we show energy-accuracy tradeoff provides clear guidance to achieve efficient realization of trained DNNs. We also observe significant implementation cost reductions with up to 11.50X in energy and 12.30X in memory while the relative performance (accuracy) loss is negligible. We also discuss and validate strategies to apply our procedure to multiple layers in a DNN in order to maximize the compression ratio of the network after simplification subject to an overall budget in degradation of accuracy across all the simplified layers.

6 August 2019 in IEEE Transactions on Very Large Scale Integration (VLSI) Systems
Verified
Journal Article
Closed access
WZWei Zeng
Wei Zeng
BZBoyu Zhang
Boyu Zhang
ADAzadeh Davoodi
Azadeh Davoodi
This paper is the first to analyze the security of split manufacturing using machine learning (ML), based on data collected from layouts provided by industry, with eight routing metal layers and significant variation in wire size and routing congestion across the layers. Many types of layout features are considered in our ML model, including those obtained from placement, routing, and cell sizes. Since the runtime cost of our basic ML procedure becomes prohibitively large for lower layers, we propose novel techniques to make it scalable with little sacrifice in the effectiveness of the attack. Moreover, we further improve the performance in the top routing layer by making use of higher quality training samples and by exploiting the routing convention. We also propose a validation-based proximity attack procedure, which generally outperforms our recent prior work. In the experiments, we analyze the ranking of the features used in our ML model and show how features vary in importance when moving to the lower layers. We provide comprehensive evaluation and comparison of our model with different configurations and demonstrate dramatically better performance of attacks compared to the prior work.

28 February 2020 in IEEE Design & Test
Verified
Journal Article
Closed access
BZBoyu Zhang
Boyu Zhang
ADAzadeh Davoodi
Azadeh Davoodi
YHYu Hen Hu
Yu Hen Hu

1 March 2017
Verified
Conference Paper
Closed access
BZBoyu Zhang
Boyu Zhang
ADAzadeh Davoodi
Azadeh Davoodi

31 October 2018 in arXiv
Verified
Preprint
Open access
PDF file
BZBoyu Zhang
Boyu Zhang
ADAzadeh Davoodi
Azadeh Davoodi
YU HEN HUYU HEN HU
YU HEN HU
The ability to customize a trained Deep Neural Network (DNN) locally using user-specific data may greatly enhance user experiences, reduce development costs, and protect user's privacy. In this work, we propose to incorporate a novel Mixture of Experts (MOE) approach to accomplish this goal. This architecture comprises of a Global Expert (GE), a Local Expert (LE) and a Gating Network (GN). The GE is a trained DNN developed on a large training dataset representative of many potential users. After deployment on an embedded edge device, GE will be subject to customized, user-specific data (e.g., accent in speech) and its performance may suffer. This problem may be alleviated by training a local DNN (the local expert, LE) on a small size customized training data to correct the errors made by GE. A gating network then will be trained to determine whether an incoming data should be handled by GE or LE. Since the customized dataset is in general very small, the cost of training LE and GN would be much lower than that of re-training of GE. The training of LE and GN thus can be performed at local device, properly protecting the privacy of customized training data. In this work, we developed a prototype MOE architecture for handwritten alphanumeric character recognition task. We use EMNIST as the generic dataset, LeNet5 as GE, and handwritings of 10 users as the customized dataset. We show that with the LE and GN, the classification accuracy is significantly enhanced over the customized dataset with almost no degradation of accuracy over the generic dataset. In terms of energy and network size, the overhead of LE and GN is around 2.5% compared to those of GE.