Turbo-smt: Accelerating coupled sparse matrix-tensor factorizations by 200x

How can we correlate the neural activity in the human brain as it responds to typed words, with properties of these terms (like ‘edible’, ‘fits in hand’)? In short, we want to find latent variables, that jointly explain both the brain activity, as well as the behavioral responses. This is one of many settings of the Coupled Matrix-Tensor Factorization (CMTF) problem.

Can we accelerate any CMTF solver, so that it runs within a few minutes instead of tens of hours to a day, while maintaining good accuracy? We introduce Turbo-SMT, a meta-method capable of doing exactly that: it boosts the performance of any CMTF algorithm, by up to 200x, along with an up to 65 fold increase in sparsity, with comparable accuracy to the baseline.

We apply Turbo-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. Turbo-SMT is able to find meaningful latent variables, as well as to predict brain activity with competitive accuracy.

Evaluating the use of testate amoebae for palaeohydrological reconstruction in permafrost peatlands

The melting of high-latitude permafrost peatlands is a major concern due to a potential positive feedback on global climate change. We examine the ecology of testate amoebae in permafrost peatlands, based on sites in Sweden (~ 200 km north of the Arctic Circle). Multivariate statistical analysis confirms that water-table depth and moisture content are the dominant controls on the distribution of testate amoebae, corroborating the results from studies in mid-latitude peatlands. We present a new testate amoeba-based water table transfer function and thoroughly test it for the effects of spatial autocorrelation, clustered sampling design and uneven sampling gradients. We find that the transfer function has good predictive power; the best-performing model is based on tolerance-downweighted weighted averaging with inverse deshrinking (performance statistics with leave-one-out cross validation: R2 = 0.87, RMSEP = 5.25 cm). The new transfer function was applied to a short core from Stordalen mire, and reveals a major shift in peatland ecohydrology coincident with the onset of the Little Ice Age (c. AD 1400). We also applied the model to an independent contemporary dataset from Stordalen and find that it outperforms predictions based on other published transfer functions. The new transfer function will enable palaeohydrological reconstruction from permafrost peatlands in Northern Europe, thereby permitting greatly improved understanding of the long-term ecohydrological dynamics of these important carbon stores as well as their responses to recent climate change.

Ecology of testate amoebae in an Amazonian peatland and development of a transfer function for palaeohydrological reconstruction

Tropical peatlands represent globally important carbon sinks with a unique biodiversity and are currently threatened by climate change and human activities. It is now imperative that proxy methods are developed to understand the ecohydrological dynamics of these systems and for testing peatland development models. Testate amoebae have been used as environmental indicators in ecological and palaeoecological studies of peatlands, primarily in ombrotrophic Sphagnum-dominated peatlands in the mid- and high-latitudes. We present the first ecological analysis of testate amoebae in a tropical peatland, a nutrient-poor domed bog in western (Peruvian) Amazonia. Litter samples were collected from different hydrological microforms (hummock to pool) along a transect from the edge to the interior of the peatland. We recorded 47 taxa from 21 genera. The most common taxa are Cryptodifflugia oviformis, Euglypha rotunda type, Phryganella acropodia, Pseudodifflugia fulva type and Trinema lineare. One species found only in the southern hemisphere, Argynnia spicata, is present. Arcella spp., Centropyxis aculeata and Lesqueresia spiralis are indicators of pools containing standing water. Canonical correspondence analysis and non-metric multidimensional scaling illustrate that water table depth is a significant control on the distribution of testate amoebae, similar to the results from mid- and high-latitude peatlands. A transfer function model for water table based on weighted averaging partial least-squares (WAPLS) regression is presented and performs well under cross-validation (r 2apparent=0.76,RMSE=4.29;r2jack=0.68,RMSEP=5.18. The transfer function was applied to a 1-m peat core, and sample-specific reconstruction errors were generated using bootstrapping. The reconstruction generally suggests near-surface water tables over the last 3,000 years, with a shift to drier conditions at c. cal. 1218-1273 AD

A LEED STRUCTURAL STUDY OF THE CO INDUCED RECONSTRUCTION OF PD(110) - EVIDENCE FOR A MISSING ROW STRUCTURE

Molecularly adsorbed CO on Pd{110} has been shown (R. Raval et al., Chem. Phys. Lett. 167 (1990) 391, ref. [1]) to induce a substantial reconstruction of the surface in the coverage range 0.3 <theta less-than-or-equal-to 0.75. Throughout this coverage range, the adsorbate-covered reconstructed surface exhibits a (4 x 2) LEED pattern. However, the exact nature of the reconstruction remains uncertain. We have conducted a LEED I(E) "fingerprinting" analysis of the CO/Pd{110}-(4 x 2) structure in order to establish the type of reconstruction induced in the metal surface. This study shows that the LEED I(E) profiles of the integral order and appropriate half-order beams of the CO/Pd{110}-(4 x 2) pattern closely resemble the I(E) profiles theoretically calculated for a Pd{110}-(1 x 2) missing-row structure. Additionally, there is a strong resemblance to the experimental LEED I(E) profiles for the Cs/Pd{110}-(1 x 2) structure which has also been shown to exhibit the missing-row structure. On the basis of this evidence we conclude that the CO/Pd{110}-(4 x 2) LEED pattern arises from a missing-row reconstruction of the Pd{110} surface which gives rise to a strong underlying (1 x 2) pattern plus a poorly ordered CO overlayer which produces weak, diffuse fourth-order spots in the LEED pattern.

Beyond Pietism and Prosperity: Religious Resources for Reconstruction and Reconciliation in Zimbabwe

It is often assumed that charismatic Christianity in Africa promotes either a pietist withdrawal from social and political concerns, or a preoccupation with gaining individual health and wealth (the prosperity gospel). This research presents an alternative vision of the role of charismatic Christianity in Zimbabwe. Drawing on an ethnographic case study of a charismatic congregation, it analyzes how these Christians are drawing links between spirituality and social action. This congregation is developing an egalitarian conception of power, promoting service to the poor, and using biblical discourses to support their actions. This can be understood as part of a wider process in which Zimbabwean Christians are using religious resources to develop a vision for reconstruction and reconciliation. This article points to further areas in which the churches could use their public position to raise sensitive issues, including how to deal with the past and heal relationships between previously antagonistic groups.

The heart of a dragon: 3D anatomical reconstruction of the 'scaly-foot gastropod' (Mollusca: Gastropoda: Neomphalina) reveals its extraordinary circulatory system

Introduction: The 'scaly-foot gastropod' (Chrysomallon squamiferum Chen et al., 2015) from deep-sea hydrothermal vent ecosystems of the Indian Ocean is an active mobile gastropod occurring in locally high densities, and it is distinctive for the dermal scales covering the exterior surface of its foot. These iron-sulfide coated sclerites, and its nutritional dependence on endosymbiotic bacteria, are both noted as adaptations to the extreme environment in the flow of hydrogen sulfide. We present evidence for other adaptations of the 'scaly-foot gastropod' to life in an extreme environment, investigated through dissection and 3D tomographic reconstruction of the internal anatomy.

Results: Our anatomical investigations of juvenile and adult specimens reveal a large unganglionated nervous system, a simple and reduced digestive system, and that the animal is a simultaneous hermaphrodite. We show that Chrysomallon squamiferum relies on endosymbiotic bacteria throughout post-larval life. Of particular interest is the circulatory system: Chrysomallon has a very large ctenidium supported by extensive blood sinuses filled with haemocoel. The ctenidium provides oxygen for the host but the circulatory system is enlarged beyond the scope of other similar vent gastropods. At the posterior of the ctenidium is a remarkably large and well-developed heart. Based on the volume of the auricle and ventricle, the heart complex represents approximately 4 % of the body volume. This proportionally giant heart primarily sucks blood through the ctenidium and supplies the highly vascularised oesophageal gland. Thus we infer the elaborate cardiovascular system most likely evolved to oxygenate the endosymbionts in an oxygen poor environment and/or to supply hydrogen sulfide to the endosymbionts.

Conclusions: This study exemplifies how understanding the autecology of an organism can be enhanced by detailed investigation of internal anatomy. This gastropod is a large and active species that is abundant in its hydrothermal vent field ecosystem. Yet all of its remarkable features-protective dermal sclerites, circulatory system, high fecundity-can be viewed as adaptations beneficial to its endosymbiont microbes. We interpret these results to show that, as a result of specialisation to resolve energetic needs in an extreme chemosynthetic environment, this dramatic dragon-like species has become a carrying vessel for its bacteria.

A multi-proxy chronological reconstruction of palaeosalinity in the eastern Limfjord, Denmark

The Jutland peninsula in northern Denmark is home to the Limfjord, one of the largest estuarine bodies of water in the region. Human inhabitance of the Limfjord’s surrounding coastlines stretches back further than 7,800 cal BP, with anthropogenic influence on the landscape beginning approximately 6,000 cal BP. Understanding how the Limfjord as a system has changed throughout time is useful in comprehending subsistence patterns and anthropogenic influence. This research is part of a larger project aimed at discerning subsistence patterns and environmental change in the region. Following the Younger Dryas, as the Fennoscandian ice sheet began to melt, Denmark experienced isostatic rebound, which contributed to the complex sea level history in the region. Between ice melt and isostatic rebound, the Jutland peninsula experienced many transgression and regression events. Connections to surrounding seas have shifted throughout time, with most attention focused on the western connection of the Limfjord with the North Sea, which has experienced numerous closures and subsequent re-openings throughout the Holocene. Furthermore, the Limfjord-North Sea connection has been the focal point of research because of the west to east water flow in the system, and the present day higher salinity in the west compared to the east. Little to no consideration has been paid to the influence of the Kattegat and Baltic on the Limfjord until now. A 10m sediment core was taken from Sebbersund (near Nibe, Limfjord), along the connection between the Limfjord and the Kattegat in the east to understand how the eastern part of the system has changed and differed from changes observed in the west. The Sebbersund sequence spans a majority of the Holocene, from 9600 cal BP to 1030 cal BP, determined via radiocarbon dating of terrestrial macrofossils and bulk sediment. Over this time period palaeoenvironmental conditions were reconstructed through the use of geochemical analyses (13C, 15N, C:N), physical sediment analyses, dinoflagellate cyst abundances and molluscan analyses. apart from two instances of low salinity, one at the top and one at the bottom of the core, the sequence has a strong marine signal for a majority of the Holocene. Radiocarbon dating of bulk sediment samples showed the presence of old carbon in the system, creating an age offset between 1,300 ± 200 and 2,800 ± 200 calibrated 14C years compared to the age-depth curve based on the terrestrial macrofossils. This finding, along with the strong marine influence in the system, discerned through geochemical data, dinoflagellate cyst and mollusc counts, is important for obtaining accurate radiocarbon ages in the region and stresses the importance of understanding both the marine and freshwater reservoir effects. The marine dominance in the eastern Limfjord differs from the west, which is characterized by a number of freshwater events when the North Sea connection was closed off during the Holocene. The eastern connection was open to the Kattegat throughout a large portion of the Holocene, with influx of open ocean water entering the system during periods of higher sea level.

Brightness variation distributions among main belt asteroids from sparse light curve sampling with Pan-STARRS 1

The rotational state of asteroids is controlled by various physical mechanisms including collisions, internal damping and the Yarkovsky-O'Keefe-Radzievskii-Paddack (YORP) effect. We have analysed the changes in magnitude between consecutive detections of ∼ 60,000 asteroids measured by the PanSTARRS 1 survey during its first 18 months of operations. We have attempted to explain the derived brightness changes physically and through the application of a simple model. We have found a tendency toward smaller magnitude variations with decreasing diameter for objects of 1 < D < 8 km. Assuming the shape distribution of objects in this size range to be independent of size and composition our model suggests a population with average axial ratios 1: 0.85 ± 0.13: 0.71 ± 0.13, with larger objects more likely to have spin axes perpendicular to the orbital plane.

Sparse multi-carrier index keying OFDM with index separation over correlated sub-carriers

In this paper, we propose a sparse multi-carrier index keying (MCIK) method for orthogonal frequency division multiplexing (OFDM) system, which uses the indices of sparse sub-carriers to transmit the data, and improve the performance
of signal detection in highly correlated sub-carriers. Although a receiver is able to exploit a power gain with precoding in OFDM, the sensitivity of the signal detection is usually high as the orthogonality is not retained in highly dispersive
environments. To overcome this, we focus on developing the trade-off between the sparsity of the MCIK, correlation, and performances, analyzing the average probability of the error propagation imposed by incorrect index detection over highly correlated sub-carriers. In asymptotic cases, we are able to see how sparsity of MCIK should be designed in order to perform superior to the classical OFDM system. Based on this feature, sparse MCIK based OFDM is a better choice for low detection errors in highly correlated sub-carriers.

Learning from the QUEST multicentre feasibility randomization trials in breast reconstruction after mastectomy

BACKGROUND: Breast reconstruction aims to improve health-related quality of life after mastectomy. However, evidence guiding patients and surgeons in shared decision-making concerning the optimal type or timing of surgery is lacking.

METHODS: QUEST comprised two parallel feasibility phase III randomized multicentre trials to assess the impact of the type and timing of latissimus dorsi breast reconstruction on health-related quality of life when postmastectomy radiotherapy is unlikely (QUEST A) or highly probable (QUEST B). The primary endpoint for the feasibility phase was the proportion of women who accepted randomization, and it would be considered feasible if patient acceptability rates exceeded 25 per cent of women approached. A companion QUEST Perspectives Study (QPS) of patients (both accepting and declining trial participation) and healthcare professionals assessed trial acceptability.

RESULTS: The QUEST trials opened in 15 UK centres. After 18 months of recruitment, 17 patients were randomized to QUEST A and eight to QUEST B, with overall acceptance rates of 19 per cent (17 of 88) and 22 per cent (8 of 36) respectively. The QPS recruited 56 patients and 51 healthcare professionals. Patient preference was the predominant reason for declining trial entry, given by 47 (53 per cent) of the 88 patients approached for QUEST A and 22 (61 per cent) of the 36 approached for QUEST B. Both trials closed to recruitment in December 2012, acknowledging the challenges of achieving satisfactory patient accrual.

CONCLUSION: Despite extensive efforts to overcome recruitment barriers, it was not feasible to reach timely recruitment targets within a feasibility study. Patient preferences for breast reconstruction types and timings were common, rendering patients unwilling to enter the trial.

A Fast Algorithm for Sparse Support Vector Machines for Mobile Computing Applications

This research presents a fast algorithm for projected support vector machines (PSVM) by selecting a basis vector set (BVS) for the kernel-induced feature space, the training points are projected onto the subspace spanned by the selected BVS. A standard linear support vector machine (SVM) is then produced in the subspace with the projected training points. As the dimension of the subspace is determined by the size of the selected basis vector set, the size of the produced SVM expansion can be specified. A two-stage algorithm is derived which selects and refines the basis vector set achieving a locally optimal model. The model expansion coefficients and bias are updated recursively for increase and decrease in the basis set and support vector set. The condition for a point to be classed as outside the current basis vector and selected as a new basis vector is derived and embedded in the recursive procedure. This guarantees the linear independence of the produced basis set. The proposed algorithm is tested and compared with an existing sparse primal SVM (SpSVM) and a standard SVM (LibSVM) on seven public benchmark classification problems. Our new algorithm is designed for use in the application area of human activity recognition using smart devices and embedded sensors where their sometimes limited memory and processing resources must be exploited to the full and the more robust and accurate the classification the more satisfied the user. Experimental results demonstrate the effectiveness and efficiency of the proposed algorithm. This work builds upon a previously published algorithm specifically created for activity recognition within mobile applications for the EU Haptimap project [1]. The algorithms detailed in this paper are more memory and resource efficient making them suitable for use with bigger data sets and more easily trained SVMs.