Application Health Monitoring for Extreme-scale Resiliency using Cooperative Fault Management.

Resiliency is and will be a critical factor in determining scientific productivity on current and exascale supercomputers, and beyond. Applications oblivious to and incapable of handling transient soft and hard errors could waste supercomputing resources or, worse, yield misleading scientific insights. We introduce a novel application-driven silent error detection and recovery strategy based on application health monitoring. Our methodology uses application output that follows known patterns as indicators of an application's health, and knowledge that violation of these patterns could be indication of faults. Information from system monitors that report hardware and software health status is used to corroborate faults. Collectively, this information is used by a fault coordinator agent to take preventive and corrective measures by applying computational steering to an application between checkpoints. This cooperative fault management system uses the Fault Tolerance Backplane as a communication channel. The benefits of this framework are demonstrated with two real application case studies, molecular dynamics and quantum chemistry simulations, on scalable clusters with simulated memory and I/O corruptions. The developed approach is general and can be easily applied to other applications.

Focus prediction in digital holographic microscopy using deep convolutional neural networks.

Deep artificial neural network learning is an emerging tool in image analysis. We demonstrate its potential in the field of digital holographic microscopy by addressing the challenging problem of determining the in-focus reconstruction depth of Madin-Darby canine kidney cell clusters encoded in digital holograms. A deep convolutional neural network learns the in-focus depths from half a million hologram amplitude images. The trained network correctly determines the in-focus depth of new holograms with high probability, without performing numerical propagation. This paper reports on extensions to preliminary work published earlier as one of the first applications of deep learning in the field of digital holographic microscopy.

Space-variant video compression and processing in digital holographic microscopy sensor networks with application to potable water monitoring.

Water-related diseases affect societies in all parts of the world. Online sensors are considered a solution to the problems associated with laboratory testing in potable water. One of the most active research areas of such online sensors has been within optics. Digital holographic microscopy (DHM) has the potential to rival state-of-the-art techniques such as advanced turbidity measurement. However, its use as an online sensor is limited by the large data requirements typical for digital holographic video. In this paper, we provide a solution that permits DHM to be applied to a whole class of online remote sensor networks, of which potable water analysis is one example. The designed sensors incorporate a novel space-variant quantization algorithm to preprocess each frame of a video sequence before transmission over a network. The system satisfies the generally accepted requirements of an online system: automated, near real-time, and operating in a real environment. To verify the effectiveness of the design, we implemented and evaluated it in an active potable water facility.

HER2 challenge contest: a detailed assessment of automated HER2 scoring algorithms in whole slide images of breast cancer tissues.

AIMS: Evaluating expression of the human epidermal growth factor receptor 2 (HER2) by visual examination of immunohistochemistry (IHC) on invasive breast cancer (BCa) is a key part of the diagnostic assessment of BCa due to its recognized importance as a predictive and prognostic marker in clinical practice. However, visual scoring of HER2 is subjective, and consequently prone to interobserver variability. Given the prognostic and therapeutic implications of HER2 scoring, a more objective method is required. In this paper, we report on a recent automated HER2 scoring contest, held in conjunction with the annual PathSoc meeting held in Nottingham in June 2016, aimed at systematically comparing and advancing the state-of-the-art artificial intelligence (AI)-based automated methods for HER2 scoring. METHODS AND RESULTS: The contest data set comprised digitized whole slide images (WSI) of sections from 86 cases of invasive breast carcinoma stained with both haematoxylin and eosin (H&E) and IHC for HER2. The contesting algorithms predicted scores of the IHC slides automatically for an unseen subset of the data set and the predicted scores were compared with the 'ground truth' (a consensus score from at least two experts). We also report on a simple 'Man versus Machine' contest for the scoring of HER2 and show that the automated methods could beat the pathology experts on this contest data set. CONCLUSIONS: This paper presents a benchmark for comparing the performance of automated algorithms for scoring of HER2. It also demonstrates the enormous potential of automated algorithms in assisting the pathologist with objective IHC scoring.

Multiwavefront digital holographic television.

This paper presents the full technology chain supporting wide angle digital holographic television from holographic capture of real world objects/scenes to holographic display with an extended viewing angle. The data are captured with multiple CCD cameras located around an object. The display system is based on multiple tilted spatial light modulators (SLMs) arranged in a circular configuration. The capture-display system is linked by a holographic data processing module, which allows for significant decoupling of the capture and display systems. The presented experimental results, based on the reconstruction of real world, variable in time scenes, illustrates imaging dynamics, viewing angle and quality.

Enhancement of three-dimensional perception of numerical hologram reconstructions of real-world objects by motion and stereo.

We investigated the question of how the perception of three-dimensional information reconstructed numerically from digital holograms of real-world objects, and presented on conventional displays, depends on motion and stereoscopic presentation. Perceived depth in an adjustable random pattern stereogram was matched to the depth in hologram reconstructions. The objects in holograms were a microscopic biological cell and a macroscopic metal coil. For control, we used real physical objects in additional to hologram reconstructions of real objects. Stereoscopic presentation increased perceived depth substantially in comparison to non-stereoscopic presentation. When stereoscopic cues were weak or absent e.g. because of blur, motion increased perceived depth considerably. However, when stereoscopic cues were strong, the effect of motion was small. In conclusion, for the maximization of perceived three-dimensional information of holograms on conventional displays, it seems highly beneficial to use the combination of motion and stereoscopic presentation.

Calculating depth maps from digital holograms using stereo disparity.

Depth extraction is an important aspect of three-dimensional (3D) image processing with digital holograms and an essential step in extended focus imaging and metrology. All available depth extraction techniques with macroscopic objects are based on variance; however, the effectiveness of this is object dependent. We propose to use disparity between corresponding points in intensity reconstructions to determine depth. Our method requires a single hologram of a scene, from which we reconstruct two different perspectives. In the reconstruction the phase information is not needed, which makes this method useful for in-line digital holography. To our knowledge disparity based 3D image processing has never been proposed before for digital holography.

Multi-heuristic dynamic task allocation using genetic algorithms in a heterogeneous distributed system.

We present a multi-heuristic evolutionary task allocation algorithm to dynamically map tasks to processors in a heterogeneous distributed system. It utilizes a genetic algorithm, combined with eight common heuristics, in an effort to minimize the total execution time. It operates on batches of unmapped tasks and can preemptively remap tasks to processors. The algorithm has been implemented on a Java distributed system and evaluated with a set of six problems from the areas of bioinformatics, biomedical engineering, computer science and cryptography. Experiments using up to 150 heterogeneous processors show that the algorithm achieves better efficiency than other state-of-the-art heuristic algorithms.

Synthesis and display of dynamic holographic 3D scenes with real-world objects.

A 3D scene is synthesized combining multiple optically recorded digital holograms of different objects. The novel idea consists of compositing moving 3D objects in a dynamic 3D scene using a process that is analogous to stop-motion video. However in this case the movie has the exciting attribute that it can be displayed and observed in 3D. We show that 3D dynamic scenes can be projected as an alternative to complicated and heavy computations needed to generate realistic-looking computer generated holograms. The key tool for creating the dynamic action is based on a new concept that consists of a spatial, adaptive transformation of digital holograms of real-world objects allowing full control in the manipulation of the object's position and size in a 3D volume with very high depth-of-focus. A pilot experiment to evaluate how viewers perceive depth in a conventional single-view display of the dynamic 3D scene has been performed.

Statistical investigation of the double random phase encoding technique.

The amplitude-encoding case of the double random phase encoding technique is examined by defining a cost function as a metric to compare an attempted decryption against the corresponding original input image. For the case when a cipher-text pair has been obtained and the correct decryption key is unknown, an iterative attack technique can be employed to ascertain the key. During such an attack the noise in the output field for an attempted decryption can be used as a measure of a possible decryption key's correctness. For relatively small systems, i.e., systems involving fewer than 5x5 pixels, the output decryption of every possible key can be examined to evaluate the distribution of the keys in key space in relation to their relative performance when carrying out decryption. However, in order to do this for large systems, checking every single key is currently impractical. One metric used to quantify the correctness of a decryption key is the normalized root mean squared (NRMS) error. The NRMS is a measure of the cumulative intensity difference between the input and decrypted images. We identify a core term in the NRMS, which we refer to as the difference parameter, d. Expressions for the expected value (or mean) and variance of d are derived in terms of the mean and variance of the output field noise, which is shown to be circular Gaussian. These expressions assume a large sample set (number of pixels and keys). We show that as we increase the number of samples used, the decryption error obeys the statistically predicted characteristic values. Finally, we corroborate previously reported simulations in the literature by using the statistically derived expressions.

Three dimensional digital holographic profiling of micro-fibers.

A method to measure the size, orientation, and location of opaque micro-fibers using digital holography is presented. The method involves the recording of a digital hologram followed by reconstruction at different depths. A novel combination of automated image analysis and statistical techniques, applied on the intensity of reconstructed digital holograms is used to accurately determine the characteristics of the micro-fibers. The performance of the proposed method is verified with a single fiber of known length and orientation. The potential of the method for measurement of fiber length is further demonstrated through its application to a suspension of fibers in a liquid medium.

Introducing secure modes of operation for optical encryption.

We analyze optical encryption systems using the techniques of conventional cryptography. All conventional block encryption algorithms are vulnerable to attack, and often they employ secure modes of operation as one way to increase security. We introduce the concept of conventional secure modes to optical encryption and analyze the results in the context of known conventional and optical attacks. We consider only the optical system "double random phase encoding," which forms the basis for a large number of optical encryption, watermarking, and multiplexing systems. We consider all attacks proposed to date in one particular scenario. We analyze only the mathematical algorithms themselves and do not consider the additional security that arises from employing these algorithms in physical optical systems.

Extended focused imaging for digital holograms of macroscopic three-dimensional objects.

When a digital hologram is reconstructed, only points located at the reconstruction distance are in focus. We have developed a novel technique for creating an in-focus image of the macroscopic objects encoded in a digital hologram. This extended focused image is created by combining numerical reconstructions with depth information extracted by using our depth-from-focus algorithm. To our knowledge, this is the first technique that creates extended focused images of digital holograms encoding macroscopic objects. We present results for digital holograms containing low- and high-contrast macroscopic objects.

Role of phase key in the double random phase encoding technique: an error analysis.

We perform a numerical analysis of the double random phase encryption-decryption technique to determine how, in the case of both amplitude and phase encoding, the two decryption keys (the image- and Fourier-plane keys) affect the output gray-scale image when they are in error. We perform perfect encryption and imperfect decryption. We introduce errors into the decrypting keys that correspond to the use of random distributions of incorrect pixel values. We quantify the effects that increasing amounts of error in the image-plane key, the Fourier-plane key, and both keys simultaneously have on the decrypted image. Quantization effects are also examined.

Low memory distributed reconstruction of large digital holograms.

We present a parallel implementation of the Fresnel transform suitable for reconstructing large digital holograms. Our method has a small memory footprint and utilizes the spare resources of a distributed set of desktop PCs connected by a network. We show how we parallelize the Fresnel transform and discuss how it is constrained by computer and communication resources. Finally, we demonstrate how a 4.3 gigapixel digital hologram can be reconstructed and how the efficiency of the method changes for different memory and processor configurations.

Noninterferometric phase retrieval using a fractional Fourier system.

The signal extraction method based on intensity measurements in two close fractional Fourier domains is examined by using the phase space formalism. The fractional order separation has a lower bound and an upper bound that depend on the signal at hand and the noise in the optical system used for measurement. On the basis of a theoretical analysis, it is shown that for a given optical system a judicious choice of fractional order separation requires some a priori knowledge of the signal bandwidth. We also present some experimental results in support of the analysis.

Key-space analysis of double random phase encryption technique.

We perform a numerical analysis on the double random phase encryption/decryption technique. The key-space of an encryption technique is the set of possible keys that can be used to encode data using that technique. In the case of a strong encryption scheme, many keys must be tried in any brute-force attack on that technique. Traditionally, designers of optical image encryption systems demonstrate only how a small number of arbitrary keys cannot decrypt a chosen encrypted image in their system. However, this type of demonstration does not discuss the properties of the key-space nor refute the feasibility of an efficient brute-force attack. To clarify these issues we present a key-space analysis of the technique. For a range of problem instances we plot the distribution of decryption errors in the key-space indicating the lack of feasibility of a simple brute-force attack.

Compression defects in different reconstructions from phase-shifting digital holographic data.

Three principal strategies for the compression of phase-shifting digital holograms (interferogram domain-, hologram domain-, and reconstruction domain-based strategies) are reviewed and their effects in the reconstruction domain are investigated. Images of the reconstructions are provided to visually compare the performances of the methods. In addition to single reconstructions the compression effects on different depth reconstructions and reconstructions corresponding to different viewing angles are investigated so that a range of the 3D aspects of the holograms may be considered. Although comparable at low compression rates, it is found that depth and perspective information is degraded in different ways with the different techniques at high compression rates. A hologram of an object with sufficient details at different depths is used so that both parallax and depth effects can be illustrated.

Histogram approaches for lossy compression of digital holograms of three-dimensional objects.

We present a novel nonuniform quantization compression technique-histogram quantization-for digital holograms of 3-D real-world objects. We exploit a priori knowledge of the distribution of the values in our data. We compare this technique to another histogram based approach: a modified version of Max's algorithm that has been adapted in a straight-forward manner to complex-valued 2-D signals. We conclude the compression procedure by applying lossless techniques to our quantized data. We demonstrate improvements over previous results obtained by applying uniform and nonuniform quantization techniques to the hologram data.

MultiPhyl: a high-throughput phylogenomics webserver using distributed computing.

With the number of fully sequenced genomes increasing steadily, there is greater interest in performing large-scale phylogenomic analyses from large numbers of individual gene families. Maximum likelihood (ML) has been shown repeatedly to be one of the most accurate methods for phylogenetic construction. Recently, there have been a number of algorithmic improvements in maximum-likelihood-based tree search methods. However, it can still take a long time to analyse the evolutionary history of many gene families using a single computer. Distributed computing refers to a method of combining the computing power of multiple computers in order to perform some larger overall calculation. In this article, we present the first high-throughput implementation of a distributed phylogenetics platform, MultiPhyl, capable of using the idle computational resources of many heterogeneous non-dedicated machines to form a phylogenetics supercomputer. MultiPhyl allows a user to upload hundreds or thousands of amino acid or nucleotide alignments simultaneously and perform computationally intensive tasks such as model selection, tree searching and bootstrapping of each of the alignments using many desktop machines. The program implements a set of 88 amino acid models and 56 nucleotide maximum likelihood models and a variety of statistical methods for choosing between alternative models. A MultiPhyl webserver is available for public use at: http://www.cs.nuim.ie/distributed/multiphyl.php.