No abstract

We propose ApproxHPVM, a compiler IR and system designed to enable accuracy-aware performance and energy tuning on heterogeneous systems with multiple compute units and approximation methods. ApproxH-PVM automatically translates end-to-end application-level quality metrics into accuracy requirements for individual operations. ApproxHPVM uses a hardware-agnostic accuracy-tuning phase to do this translation that provides greater portability across heterogeneous hardware platforms and enables future capabilities like accuracy-aware dynamic scheduling and design space exploration. ApproxHPVM incorporates three main components: (a) a compiler IR with hardware-agnostic approximation metrics, (b) a hardware-agnostic accuracy-tuning phase to identify error-tolerant computations, and (c) an accuracy-aware hardware scheduler that maps error-tolerant computations to approximate hardware components. As ApproxHPVM does not incorporate any hardware-specific knowledge as part of the IR, it can serve as a portable virtual ISA that can be shipped to all kinds of hardware platforms. We evaluate our framework on nine benchmarks from the deep learning domain and five image processing benchmarks. Our results show that our framework can offload chunks of approximable computations to special-purpose accelerators that provide significant gains in performance and energy, while staying within user-specified application-level quality metrics with high probability. Across the 14 benchmarks, we observe from 1-9x performance speedups and 1.1-11.3x energy reduction for very small reductions in accuracy. CCS Concepts: • Software and its engineering → Compilers.

This article reports performance characteristics of the population bioequivalence (PBE) statistical test recommended by the US Food and Drug Administration (FDA) for orally inhaled products. A PBE Working Group of the International Pharmaceutical Aerosol Consortium on Regulation and Science (IPAC-RS) assembled and considered a database comprising delivered dose measurements from 856 individual batches across 20 metered dose inhaler products submitted by industry. A review of the industry dataset identified variability between batches and a systematic lifestage effect that was not included in the FDA-prescribed model for PBE. A simulation study was designed to understand PBE performance when factors identified in the industry database were present. Neglecting between-batch variability in the PBE model inflated errors in the equivalence conclusion: (i) The probability of incorrectly concluding equivalence (type I error) often exceeded 15% for non-zero between-batch variability, and (ii) the probability of incorrectly rejecting equivalence (type II error) for identical products approached 20% when product and between-batch variabilities were high. Neglecting a systematic through-life increase in the PBE model did not substantially impact PBE performance for the magnitude of lifestage effect considered. Extreme values were present in 80% of the industry products considered, with low-dose extremes having a larger impact on equivalence conclusions. The dataset did not support the need for log-transformation prior to analysis, as requested by FDA. Log-transformation resulted in equivalence conclusions that depended on the direction of product mean differences. These results highlight a need for further refinement of in vitro equivalence methodology.

With the increase in the popularity of using huge image databases in various image retrieval applications, a need arises to develop an efficient, robust and automatic system which provides output in the form of similar images with respect to the input or query image. In this paper, three Image Retrieval Systems using very basic but novel descriptors, which can retrieve black and white (binary) image, grayscale image and color image from the binary, grayscale and color image databases, respectively, were implemented. It should be noted that the features used for the binary image and grayscale image may not be efficiently applicable to the color images. Obtained results indicate the same and it additionally shows that irrespective of numbers of images in database, developed Image Retrieval Systems retrieve images in very short period of time with high accuracy. Accuracy in color image retrieval is less compared to binary and grayscale image retrieval. So, it creates a need to design highly effective features for the color image retrieval.

No abstract