Behavioral Non-portability in Scientific Numeric Computing

Gu, Yijia; Wahl, Thomas; Bayati, Mahsa; Leeser, Miriam

doi:10.1007/978-3-662-48096-0_43

Cited by 9 publications

(6 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The average CS reconstruction time of 1.3 minutes on the GPU suggests that this approach may be suitable for in-line reconstruction on the scanner in the clinical environment. The slight difference in image intensity that we found between the 2 computer hardware configurations may be because the CS algorithm was implemented using sequential coding on the CPU and parallel coding on the GPU leading to a different order of operations, and therefore, unequal results (16).…”

Section: Discussionmentioning

confidence: 94%

Accelerated whole‐heart MR angiography using a variable‐density poisson‐disc undersampling pattern and compressed sensing reconstruction

Moghari

Uecker

Roujol

et al. 2017

Magnetic Resonance in Med

View full text Add to dashboard Cite

show abstract

Section: Discussionmentioning

confidence: 94%

Accelerated whole‐heart MR angiography using a variable‐density poisson‐disc undersampling pattern and compressed sensing reconstruction

Moghari

Uecker

Roujol

et al. 2017

Magnetic Resonance in Med

View full text Add to dashboard Cite

show abstract

“…This small mean difference is potentially due to the fact that different order of operations could lead to different results in floating-point arithmetic. In our study, the PICS stage was implemented using sequential coding on the CPU and parallel coding on the GPU, which yielded to different order of operations and therefore unequal results [17].…”

Section: Discussionmentioning

confidence: 99%

Cardiac MRI compressed sensing image reconstruction with a graphics processing unit

Sabbagh

Uecker

Powell

et al. 2016

2016 10th International Symposium on Medical Information and Communication Technology (ISMICT)

Self Cite

View full text Add to dashboard Cite

Compressed sensing (CS) magnetic resonance imaging (MRI) reconstruction reduces the scan time by undersampling the data but increases the image reconstruction time because a non-linear optimization problem must be iteratively solved to reconstruct the images. The growing demand for reducing the examination time in cardiac MRI led us to investigate opportunities to accelerate this non-linear optimization problem to facilitate the migration of CS into the clinical environment. Using 3D steady-state free precession MRI images from 5 patients, we compared the speed and output quality of CS reconstruction using central processing unit (CPU), CPU with OpenMP parallelization, and graphics processing unit (GPU) platforms. Mean reconstruction time was 13.1 ± 3.8 minutes for the CPU, 11.6 ± 3.6 minutes for the CPU with OpenMP parallelization, and 2.5 ± 0.3 minutes for the CPU with OpenMP plus GPU. GPU and CPU reconstructed image quality as assessed by image subtraction were comparable. Additional developments needed for implementation of rapid CS image reconstruction in the clinical environment are discussed.

show abstract

“…In this paper, instead of solving an optimization problem, we determine input samples that give rise to a suitable "target sensitivity". Requiring this sensitivity to be positive is not enough: differences in the compiler, the computation environment, the available hardware and other unknowns (which impact the precise semantics of floating-point arithmetic [6]) will typically cause some deviations in the output between the client's platform and the cloud. In the absence of an attack, these deviations would show up as false positives.…”

Section: Sensitive Sample Queriesmentioning

confidence: 99%

“…As discussed in Sec. 3.2, this is not quite true: due to platformdependencies of floating-point computations [6], DNN model inference is not deterministic. We address this problem using an empirical non-zero sensitivity detection threshold (Sec.…”

Section: Related Workmentioning

confidence: 99%

Sensitive Samples Revisited: Detecting Neural Network Attacks Using Constraint Solvers

Docena

Wahl

Pearce

et al. 2021

Electron. Proc. Theor. Comput. Sci.

Self Cite

View full text Add to dashboard Cite

Neural Networks are used today in numerous security-and safety-relevant domains and are, as such, a popular target of attacks that subvert their classification capabilities, by manipulating the network parameters. Prior work has introduced sensitive samples-inputs highly sensitive to parameter changes-to detect such manipulations, and proposed a gradient ascent-based approach to compute them. In this paper we offer an alternative, using symbolic constraint solvers. We model the network and a formal specification of a sensitive sample in the language of the solver and ask for a solution. This approach supports a rich class of queries, corresponding, for instance, to the presence of certain types of attacks. Unlike earlier techniques, our approach does not depend on convex search domains, or on the suitability of a starting point for the search. We address the performance limitations of constraint solvers by partitioning the search space for the solver, and exploring the partitions according to a balanced schedule that still retains completeness of the search. We demonstrate the impact of the use of solvers in terms of functionality and search efficiency, using a case study for the detection of Trojan attacks on Neural Networks.

show abstract

Behavioral Non-portability in Scientific Numeric Computing

Cited by 9 publications

References 7 publications

Accelerated whole‐heart MR angiography using a variable‐density poisson‐disc undersampling pattern and compressed sensing reconstruction

Accelerated whole‐heart MR angiography using a variable‐density poisson‐disc undersampling pattern and compressed sensing reconstruction

Cardiac MRI compressed sensing image reconstruction with a graphics processing unit

Sensitive Samples Revisited: Detecting Neural Network Attacks Using Constraint Solvers

Contact Info

Product

Resources

About