Keith M. Chugg

Keith M. Chugg

Auto-generated profile, based on publicly available information

Keith M. Chugg

University of Southern California 90089 Los Angeles CA USA

Research Focus & Metrics

Academic Career Length

21Years (2006-Present)

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

Representative Publications

View All Publications
1 January 2020 in IEEE Transactions on Computers
Verified
Journal Article
Closed access
SKSouvik Kundu
Souvik Kundu
MNMahdi Nazemi
Mahdi Nazemi
MPMassoud Pedram
Massoud Pedram
KCKeith M. Chugg
Keith M. Chugg
PBPeter Beeral
Peter Beeral
The high energy cost of processing deep convolutional neural networks impedes their ubiquitous deployment in energy-constrained platforms such as embedded systems and IoT devices. This work introduces convolutional layers with pre-defined sparse 2D kernels that have support sets that repeat periodically within and across filters. Due to the efficient storage of our periodic sparse kernels, the parameter savings can translate into considerable improvements in energy efficiency due to reduced DRAM accesses, thus promising significant improvements in the trade-off between energy consumption and accuracy for both training and inference. To evaluate this approach, we performed experiments with two widely accepted datasets, CIFAR-10 and Tiny ImageNet in sparse variants of the ResNet18 and VGG16 architectures. Compared to baseline models, our proposed sparse variants require up to ∼82% fewer model parameters with 5.6× fewer FLOPs with negligible loss in accuracy for ResNet18 on CIFAR-10. For VGG16 trained on Tiny ImageNet, our approach requires 5.8× fewer FLOPs and up to ∼83.3% fewer model parameters with a drop in top-5 (top-1) accuracy of only 1.2%(∼2.1%). We also compared the performance of our proposed architectures with that of ShuffleNet and MobileNetV2. Using similar hyperparameters and FLOPs, our ResNet18 variants yield an average accuracy improvement of ∼2.8%.

1 July 2006
Verified
Conference Paper
Closed access
THThomas R. Halford
Thomas R. Halford
KCKeith M. Chugg
Keith M. Chugg
A number of authors have recently considered iterative soft-in soft-out (SISO) decoding algorithms for classical linear block codes that utilize redundant Tanner graphs. Jiang and Narayanan presented a practically realizable algorithm that applies only to cyclic codes while Kothiyal et al. presented an algorithm that, while applicable to arbitrary linear block codes, does not imply a low-complexity implementation. This work first presents the aforementioned algorithms in a common framework and then presents a related algorithm - random redundant iterative decoding - that is both practically realizable and applicable to arbitrary linear block codes. Simulation results illustrate the successful application of the random redundant iterative decoding algorithm to the extended binary Golay code. Additionally, the proposed algorithm is shown to outperform Jiang and Narayanan's algorithm for a number of Bose-Chaudhuri-Hocquenghem (BCH) codes

1 April 2006 in Journal of Signal Processing Systems
Verified
Journal Article
Closed access
PDF file
OYOn Wa Yeung
On Wa Yeung
KCKeith M. Chugg
Keith M. Chugg

1 October 2006
Verified
Conference Paper
Closed access
DLDavid K. Lee
David K. Lee
KCKeith M. Chugg
Keith M. Chugg
Cooperative communication represents an effective way of introducing spatial diversity in wireless networks. Full performance gain of an ideal cooperative communication system requires the cooperating transmitters to form their signals such that the destination receiver is able to coherently combine the desired signals. This encompasses multiple implementation challenges. We introduce a more pragmatic approach to cooperative diversity based on modern error correction coding working with phase dithering. Modulation constrained information rate analysis for this approach is presented, and we show that the loss relative to the ideal cooperative diversity scheme is minimal. By incorporating further practical constraints, we show performance of block-oriented phase dithering within a quasi-static Rayleigh fading channel

1 July 2006
Verified
Conference Paper
Closed access
YCYuan-Sheng Cheng
Yuan-Sheng Cheng
MNMichael J. Neely
Michael J. Neely
KCKeith M. Chugg
Keith M. Chugg
We derive iterative message passing update rules for solving the bipartite maximum weighted matching problem. It is shown that if the optimal matching solution is unique, the algorithm converges to this optimal solution at a rate comparable to the algorithm of Bayati et. al. It is shown that the two algorithms are both standard messages passing, but on dual graphs of each other. Also, the algorithm presented here requires less storage. We also provide a method to use the proposed algorithm to solve the integer maximal weighted matching problem - i.e., where the optimal solution is generally not unique

1 October 2006
Verified
Conference Paper
Closed access
FPFabio Principe
Fabio Principe
KCKeith M. Chugg
Keith M. Chugg
MLMarco Luise
Marco Luise
A recent study, [1], demonstrated that iterative message passing algorithms (iMPAs) can be applied to rapidly acquire pseudo-noise (PN) sequences with low-complexity. Furthermore, a second work, [2], showed that significant benefits can be obtained using redundant graphical models, in case of linear feedback shift register (LFSR) sequences with sparse generator polynomials. Starting from these results, we address the problem of fast detection to Gold codes and LFSR sequences with dense generators. We will prove that these two aspects are closely related, and that, constructing redundant Tanner Graphs (TGs) from sparse higher-degree generator polynomials, it is possible to rapidly acquire these PN sequences at low signal-to-noise ratio (SNR), and with a low complexity. We also propose another distinct approach for acquiring Gold sequences using iterative methods based on a hierarchical model for the two LFSR generators that comprise a Gold code.