Scoup-SMT: Scalable Coupled Sparse Matrix-Tensor Factorization

How can we correlate neural activity in the human brain as it responds to words, with behavioral data expressed as answers to questions about these same words? In short, we want to find latent variables, that explain both the brain activity, as well as the behavioral responses. We show that this is an instance of the Coupled Matrix-Tensor Factorization (CMTF) problem. We propose Scoup-SMT, a novel, fast, and parallel algorithm that solves the CMTF problem and produces a sparse latent low-rank subspace of the data. In our experiments, we find that Scoup-SMT is 50-100 times faster than a state-of-the-art algorithm for CMTF, along with a 5 fold increase in sparsity. Moreover, we extend Scoup-SMT to handle missing data without degradation of performance. We apply Scoup-SMT to BrainQ, a dataset consisting of a (nouns, brain voxels, human subjects) tensor and a (nouns, properties) matrix, with coupling along the nouns dimension. Scoup-SMT is able to find meaningful latent variables, as well as to predict brain activity with competitive accuracy. Finally, we demonstrate the generality of Scoup-SMT, by applying it on a Facebook dataset (users, friends, wall-postings); there, Scoup-SMT spots spammer-like anomalies.

Container-based Cloud Virtual Machine Benchmarking

With the availability of a wide range of cloud Virtual Machines (VMs) it is difficult to determine which VMs can maximise the performance of an application. Benchmarking is commonly used to this end for capturing the performance of VMs. Most cloud benchmarking techniques are typically heavyweight - time consuming processes which have to benchmark the entire VM in order to obtain accurate benchmark data. Such benchmarks cannot be used in real-time on the cloud and incur extra costs even before an application is deployed.

In this paper, we present lightweight cloud benchmarking techniques that execute quickly and can be used in near real-time on the cloud. The exploration of lightweight benchmarking techniques are facilitated by the development of DocLite - Docker Container-based Lightweight Benchmarking. DocLite is built on the Docker container technology which allows a user-defined portion (such as memory size and the number of CPU cores) of the VM to be benchmarked. DocLite operates in two modes, in the first mode, containers are used to benchmark a small portion of the VM to generate performance ranks. In the second mode, historic benchmark data is used along with the first mode as a hybrid to generate VM ranks. The generated ranks are evaluated against three scientific high-performance computing applications. The proposed techniques are up to 91 times faster than a heavyweight technique which benchmarks the entire VM. It is observed that the first mode can generate ranks with over 90% and 86% accuracy for sequential and parallel execution of an application. The hybrid mode improves the correlation slightly but the first mode is sufficient for benchmarking cloud VMs.

SPLI is a Chemoattractant for Oral and Skin Fibroblasts

Objectives: Unlike adult dermal wounds, the oral mucosa demonstrates preferential healing characterized by rapid remodeling and re-epithelialisation, with minimal scar formation. Secretory leukocyte protease inhibitor (SLPI) is an epithelial-derived factor with potential for promoting scarless repair. The aims of this study were to: (i) investigate the directed migratory (chemotaxis) response of oral and skin fibroblasts to various concentrations of SLPI; and (ii) compare migratory speed of the two cell types. Methods: Paired oral and skin fibroblasts were seeded at 2x104 cells in six well plates containing glass coverslips, and cultured in DMEM supplemented with 10% FCS for 48hours. Following a period of serum starvation (18hours in DMEM plus 0.5% BSA), coverslips were incorporated within a Dunn chemotaxis chamber containing DMEM with 0.5% BSA +/- SLPI gradients at 0.5, 1 or 2µM concentrations. Using microscopy, the migratory behaviour of cells was digitally captured every 10mins for 18hours, traced with JCell tracking software and resulting co-ordinates statistically analysed using Mathmatica software. Results: At all concentrations SLPI was a significant chemoattractant (p<0.01) for both cell types. However, skin fibroblasts migrated significantly faster than oral cells at each SLPI concentration, with greatest effect observed at the highest dose (skin: 32.0±0.47µm/hr, oral: 13.6±0.23µm/hr). Conclusion: SLPI is a chemoattractant for both oral and skin fibroblasts, and may play an important role in fibroblast recruitment during wound healing. This work was funded by the R&D Office, N.Ireland.

DocLite: A Docker-Based Lightweight Cloud Benchmarking Tool

Existing benchmarking methods are time consuming processes as they typically benchmark the entire Virtual Machine (VM) in order to generate accurate performance data, making them less suitable for real-time analytics. The research in this paper is aimed to surmount the above challenge by presenting DocLite - Docker Container-based Lightweight benchmarking tool. DocLite explores lightweight cloud benchmarking methods for rapidly executing benchmarks in near real-time. DocLite is built on the Docker container technology, which allows a user-defined memory size and number of CPU cores of the VM to be benchmarked. The tool incorporates two benchmarking methods - the first referred to as the native method employs containers to benchmark a small portion of the VM and generate performance ranks, and the second uses historic benchmark data along with the native method as a hybrid to generate VM ranks. The proposed methods are evaluated on three use-cases and are observed to be up to 91 times faster than benchmarking the entire VM. In both methods, small containers provide the same quality of rankings as a large container. The native method generates ranks with over 90% and 86% accuracy for sequential and parallel execution of an application compared against benchmarking the whole VM. The hybrid method did not improve the quality of the rankings significantly.

The VLT-FLAMES Tarantula Survey XXI. Stellar spin rates of O-type spectroscopic binaries

Context: The initial distribution of spin rates of massive stars is a fingerprint of their elusive formation process. It also sets a key initial condition for stellar evolution and is thus an important ingredient in stellar population synthesis. So far, most studies have focused on single stars. Most O stars are, however, found in multiple systems. 

Aims: By establishing the spin-rate distribution of a sizeable sample of O-type spectroscopic binaries and by comparing the distributions of binary subpopulations with one another and with that of presumed-single stars in the same region, we aim to constrain the initial spin distribution of O stars in binaries, and to identify signatures of the physical mechanisms that affect the evolution of the spin rates of massive stars. 

Methods: We use ground-based optical spectroscopy obtained in the framework of the VLT-FLAMES Tarantula Survey (VFTS) to establish the projected equatorial rotational velocities (ν_esini) for components of 114 spectroscopic binaries in 30 Doradus. The ν_esini values are derived from the full width at half maximum (FWHM) of a set of spectral lines, using a FWHM vs. ν_esini calibration that we derive based on previous line analysis methods applied to single O-type stars in the VFTS sample. 

Results: The overall ν_esini distribution of the primary stars resembles that of single O-type stars in the VFTS, featuring a low-velocity peak (at ν_esini<200 kms^-1) and a shoulder at intermediate velocities (200 <ν_esini<300 kms^-1). The distributions of binaries and single stars, however, differ in two ways. First, the main peak at ν_esini ~ 100kms^-1 is broader and slightly shifted towards higher spin rates in the binary distribution than that of the presumed-single stars. This shift is mostly due to short-period binaries (P_orb~<10 d). Second, the ν_esini distribution of primaries lacks a significant population of stars spinning faster than 300 kms^-1, while such a population is clearly present in the single-star sample. The ν_esini distribution of binaries with amplitudes of radial velocity variation in the range of 20 to 200 kms^-1 (mostly binaries with P_orb ~ 10-1000 d and/or with q<0.5) is similar to that of single O stars below ν_esini_~<170kms^-1. 

Conclusions: Our results are compatible with the assumption that binary components formed with the same spin distribution as single stars, and that this distribution contains few or no fast-spinning stars. The higher average spin rate of stars in short-period binaries may either be explained by spin-up through tides in such tight binary systems, or by spin-down of a fraction of the presumed-single stars and long-period binaries through magnetic braking (or by a combination of both mechanisms). Most primaries and secondaries of SB2 systems with P_orb~<10 d appear to have similar rotational velocities. This is in agreement with tidal locking in close binaries where the components have similar radii. The lack of very rapidly spinning stars among binary systems supports the idea that most stars with ν_esini_~> 300kms^-1 in the single-star sample are actually spun-up post-binary interaction products. Finally, the overall similarities (low-velocity peak and intermediate-velocity shoulder) of the spin distribution of binary and single stars argue for a massive star formation process in which the initial spin is set independently of whether stars are formed as single stars or as components of a binary system.

Microscopic Mechanism and Kinetics of Ice Formation at Complex Interfaces: Zooming in on Kaolinite

Most ice in nature forms because of impurities which boost the exceedingly low nucleation rate of pure supercooled water. However, the microscopic details of ice nucleation on these substances remain largely unknown. Here, we have unraveled the molecular mechanism and the kinetics of ice formation on kaolinite, a clay mineral playing a key role in climate science. We find that the formation of ice at strong supercooling in the presence of this clay is about 20 orders of magnitude faster than homogeneous freezing. The critical nucleus is substantially smaller than that found for homogeneous nucleation and, in contrast to the predictions of classical nucleation theory (CNT), it has a strong two-dimensional character. Nonetheless, we show that CNT describes correctly the formation of ice at this complex interface. Kaolinite also promotes the exclusive nucleation of hexagonal ice, as opposed to homogeneous freezing where a mixture of cubic and hexagonal polytypes is observed.

Reversibility Iteration Method to Understand Reaction Networks and to Solve Micro-Kinetics in Heterogeneous Catalysis

Solving microkinetics of catalytic systems, which bridges microscopic processes and macroscopic reaction rates, is currently vital for understanding catalysis in silico. However, traditional microkinetic solvers possess several drawbacks that make the process slow and unreliable for complicated catalytic systems. In this paper, a new approach, the so-called reversibility iteration method (RIM), is developed to solve microkinetics for catalytic systems. Using the chemical potential notation we previously proposed to simplify the kinetic framework, the catalytic systems can be analytically illustrated to be logically equivalent to the electric circuit, and the reaction rate and coverage can be calculated by updating the values of reversibilities. Compared to the traditional modified Newton iteration method (NIM), our method is not sensitive to the initial guess of the solution and typically requires fewer iteration steps. Moreover, the method does not require arbitrary-precision arithmetic and has a higher probability of successfully solving the system. These features make it ∼1000 times faster than the modified Newton iteration method for the systems we tested. Moreover, the derived concept and the mathematical framework presented in this work may provide new insight into catalytic reaction networks.

Novel ketone diet enhances physical and cognitive performance

Ketone bodies are the most energy-efficient fuel and yield more ATP per mole of substrate than pyruvate and increase the free energy released from ATP hydrolysis. Elevation of circulating ketones via high-fat, low-carbohydrate diets has been used for the treatment of drug-refractory epilepsy and for neurodegenerative diseases, such as Parkinson's disease. Ketones may also be beneficial for muscle and brain in times of stress, such as endurance exercise. The challenge has been to raise circulating ketone levels by using a palatable diet without altering lipid levels. We found that blood ketone levels can be increased and cholesterol and triglycerides decreased by feeding rats a novel ketone ester diet: chow that is supplemented with (R)-3-hydroxybutyl (R)-3-hydroxybutyrate as 30% of calories. For 5 d, rats on the ketone diet ran 32% further on a treadmill than did control rats that ate an isocaloric diet that was supplemented with either corn starch or palm oil (P < 0.05). Ketone-fed rats completed an 8-arm radial maze test 38% faster than did those on the other diets, making more correct decisions before making a mistake (P < 0.05). Isolated, perfused hearts from rats that were fed the ketone diet had greater free energy available from ATP hydrolysis during increased work than did hearts from rats on the other diets as shown by using [(31)P]-NMR spectroscopy. The novel ketone diet, therefore, improved physical performance and cognitive function in rats, and its energy-sparing properties suggest that it may help to treat a range of human conditions with metabolic abnormalities.