Generalized Nesterov Accelerated Conjugate Gradient Algorithm for a Compressively Sampled MR Imaging Reconstruction

Li, Xiuhan; Wang, Wei; Zhu, Songsheng; Xiang, Wentao; Wu, Xiaoling

doi:10.1109/access.2020.3018446

Cited by 5 publications

(9 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…They do not require any query access to the model to craft adversarial examples. Crafting highly transferable adversarial examples for distinct architectures is still an open problem [17] and an active area of research arXiv:2304.02688v1 [cs.LG] 5 Apr 2023 [2,4,8,9,15,16,20,24,25,28]. Understanding the underlining characteristics that drive transferability provides insights into how DNNs learn generic representations.…”

Section: Rfn Sgdmentioning

confidence: 99%

“…Transferability techniques. The transferability of adversarial examples is a prolific research topic [2,4,8,9,15,16,20,24,25,28]. [28] recently categorizes of transferability techniques and recommend evaluating techniques by comparing them competitively within each category.…”

Section: Related Workmentioning

confidence: 99%

“…Model augmentation adds randomness to the weights or the architecture to avoid adversarial examples too specific to a single representation: GN [15] uses dropout or skip erosion, SGM [24] favors gradients from skip connections during the backward pass, LGV [9] collects models along the SGD trajectory during a few additional epochs with a high learning rate. Data augmentation techniques transform the inputs during the attack: DI [25] randomly resizes the input, SI [16] rescales the input, VT [23] smooths the gradients locally. Attack optimizers smooth updates during gradient ascent with momentum (MI [4]) or Nesterov accelerated gradient (NI [16]).…”

Section: Related Workmentioning

confidence: 99%

“…Data augmentation techniques transform the inputs during the attack: DI [25] randomly resizes the input, SI [16] rescales the input, VT [23] smooths the gradients locally. Attack optimizers smooth updates during gradient ascent with momentum (MI [4]) or Nesterov accelerated gradient (NI [16]).…”

Section: Related Workmentioning

confidence: 99%

“…Attack. Unless specified otherwise, we use the BIM (Basic Iterative Method)[13] which is the standard attack for transferability[2,4, 8,9,15,16,20,24,25,28]. By default, the maximum L ∞ perturbation norm ε is set to 4/255.…”

mentioning

confidence: 99%

See 4 more Smart Citations

LGV: Boosting Adversarial Example Transferability from Large Geometric Vicinity

Gubri

Cordy

Papadakis

et al. 2022

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Transferability is the property of adversarial examples to be misclassified by other models than the surrogate model for which they were crafted. Previous research has shown that transferability is substantially increased when the training of the surrogate model has been early stopped. A common hypothesis to explain this is that the later training epochs are when models learn the non-robust features that adversarial attacks exploit. Hence, an early stopped model is more robust (hence, a better surrogate) than fully trained models.We demonstrate that the reasons why early stopping improves transferability lie in the side effects it has on the learning dynamics of the model. We first show that early stopping benefits transferability even on models learning from data with non-robust features. We then establish links between transferability and the exploration of the loss landscape in the parameter space, on which early stopping has an inherent effect. More precisely, we observe that transferability peaks when the learning rate decays, which is also the time at which the sharpness of the loss significantly drops.This leads us to propose RFN, a new approach for transferability that minimizes loss sharpness during training in order to maximize transferability. We show that by searching for large flat neighborhoods, RFN always improves over early stopping (by up to 47 points of transferability rate) and is competitive to (if not better than) strong state-of-theart baselines.

show abstract

Section: Rfn Sgdmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

mentioning

confidence: 99%

See 3 more Smart Citations

LGV: Boosting Adversarial Example Transferability from Large Geometric Vicinity

Gubri

Cordy

Papadakis

et al. 2022

Lecture Notes in Computer Science

View full text Add to dashboard Cite

show abstract

On the Convergence of an Adaptive Momentum Method for Adversarial Attacks

Long,

Tao,

et al. 2024

AAAI

View full text Add to dashboard Cite

Adversarial examples are commonly created by solving a constrained optimization problem, typically using sign-based methods like Fast Gradient Sign Method (FGSM). These attacks can benefit from momentum with a constant parameter, such as Momentum Iterative FGSM (MI-FGSM), to enhance black-box transferability. However, the monotonic time-varying momentum parameter is required to guarantee convergence in theory, creating a theory-practice gap. Additionally, recent work shows that sign-based methods fail to converge to the optimum in several convex settings, exacerbating the issue. To address these concerns, we propose a novel method which incorporates both an innovative adaptive momentum parameter without monotonicity assumptions and an adaptive step-size scheme that replaces the sign operation. Furthermore, we derive a regret upper bound for general convex functions. Experiments on multiple models demonstrate the efficacy of our method in generating adversarial examples with human-imperceptible noise while achieving high attack success rates, indicating its superiority over previous adversarial example generation methods.

show abstract

Analysis of Sparse Signal Sequences under Compressive Sampling Techniques for Different Measurement Matrices

Devendrappa¹,

Karthik²,

Ananth³

2023

EEENG

View full text Add to dashboard Cite

INTRODUCTION: A more modern, extremely applicable method for signal acquisition is compression sensing. It permits effective data sampling at a rate that is significantly lower than what the Nyquist theorem suggests. Compressive sensing has a number of benefits, including a much-reduced demand for sensory devices, a smaller memory storage need, a greater data transfer rate, and significantly lower power usage. Compressive sensing has been employed in a variety of applications because of all these benefits. Neuro-signal acquisition is a domain in which compressive sensing has applications in the medical industry METHODS: The novel methods discussed in this article are FFT-based CoSaMP (FFTCoSaMP), DCT-based CoSaMP(DCTCoSaMP) and DWT-based CoSaMP (DWTCoSaMP) based on sparse signal sequences / dictionaries by means of Transform Techniques, where CoSaMP stands for Compressive Sampling Matching Pursuit with respect to Objective Quality Assessment Algorithms like PSNR, SSIM and RMSE, where CoSaMP stands for Compressive Sampling Matching Pursuit. RESULT: DWTCoSaMP is giving the PSNR values of 40.26 db for DCTCoSaMP and FFTCoSaMP, it is 36.76 db and 34.76 db. For DWTCoSaMP, SSIM value is 0.8164, and for DCTCoSaMP and FTCoSaMP, it is 0.719 and 0.5625 respectively. CONCLUSION: Finally, for DWTCoSaMP, RMSE value is 0.442, and for DCTCoSaMP and FFTCoSaMP, it is 0.44 and 0.4425, respectively. Among Compressed sampling techniques DWTCoSaMP, DCTCoSaMP and FFTCoSaMP discussed in this paper, DWTCoSaMP reveals the best results.

show abstract

Generalized Nesterov Accelerated Conjugate Gradient Algorithm for a Compressively Sampled MR Imaging Reconstruction

Cited by 5 publications

References 24 publications

LGV: Boosting Adversarial Example Transferability from Large Geometric Vicinity

LGV: Boosting Adversarial Example Transferability from Large Geometric Vicinity

On the Convergence of an Adaptive Momentum Method for Adversarial Attacks

Analysis of Sparse Signal Sequences under Compressive Sampling Techniques for Different Measurement Matrices

Contact Info

Product

Resources

About