ANU Research Publications

Permanent URI for this collectionhttps://hdl.handle.net/1885/26

The Australian National University's Research Publications collection is an online location for collecting, preserving and disseminating the scholarly output of the University. This service allows members of the University to share their research with the wider community. ANU Open Research accepts journal articles, conference papers, book chapters, working or technical papers and other forms of scholarly communication.

FINDING WORKS: To find a work in this collection, enter a search term into the text box and click on Search or select one of the 'Browse by' options (Author, Title, Subject, Issue Date or Type).

CONTRIBUTING WORKS: If you have an item you wish to submit, please do so through the Contribute page.

Browse

Now showing 1 - 20 of 121564

Metadata only
Improving Device-Edge Cooperative Inference of Deep Learning via 2-Step Pruning
(Institute of Electrical and Electronics Engineers Inc., 2019) Shi, Wenqi; Hou, Yunzhong; Zhou, Sheng; Niu, Zhisheng; Zhang, Yang; Geng, Lu
Deep neural networks (DNNs) are state-of-the-art solutions for many machine learning applications, and have been widely used on mobile devices. Running DNNs on resourceconstrained mobile devices often requires the help from edge servers via computation offloading. However, offloading through a bandwidth-limited wireless link is non-trivial due to the tight interplay between the computation resources on mobile devices and wireless resources. Existing studies have focused on cooperative inference where DNN models are partitioned at different neural network layers, and the two parts are executed at the mobile device and the edge server, respectively. Since the output data size of a DNN layer can be larger than that of the raw data, offloading intermediate data between layers can suffer from high transmission latency under limited wireless bandwidth. In this paper, we propose an efficient and flexible 2-step pruning framework for DNN partition between mobile devices and edge servers. In our framework, the DNN model only needs to be pruned once in the training phase where unimportant convolutional filters are removed iteratively. By limiting the pruning region, our framework can greatly reduce either the wireless transmission workload of the device or the total computation workload. A series of pruned models are generated in the training phase, from which the framework can automatically select to satisfy varying latency and accuracy requirements. Furthermore, coding for the intermediate data is added to provide extra transmission workload reduction. Our experiments show that the proposed framework can achieve up to 25.6X reduction on transmission workload, 6.01X acceleration on total computation and 4.81X reduction on end-to-end latency as compared to partitioning the original DNN model without pruning.
Metadata only
Effect of altitude on the carbon‐isotope composition of forest and grassland soils from Papua New Guinea
(1994) Bird, M. I.; Haberle, S. G.; Chivas, A. R.
The carbon‐isotope composition of both forest and grassland soils from Papua New Guinea exhibit predictable trends with increasing altitude. Soils under pure C3 vegetation (forests and alpine grasslands above 4000 m) show an increase in δ13C value with altitude paralleling the increase in δ13C value observed in plant leaves by Körner et al. [1988]. Grassland soils show a decrease in δ13C value above about 1000 m, from maximum values which are close to pure C4 values (−12 to −13‰ vs. PDB) to minimum values which are indistinguishable from pure C3 values at 3500–4000 m (∼−26‰). Within this general framework, several factors can influence the soil δ13C value at individual locations. In local forest settings, soil δ13C values will be influenced by the degree to which respired CO2 is re‐utilized during photosynthesis, the proportions of leaf and wood litter, and the degree of decomposition. In grassland settings the primary factor controlling the observed δ13C variability at any specific altitude is the amount of nongrass C3 carbon present in the sample. It is also possible that other factors, such as moisture availability, may play some role in determining the proportions of C3 and C4 grasses at any given altitude, although further work would be required to substantiate such a link. The results provide a framework within which to more accurately constrain the carbon‐isotope composition of terrestrial carbon pools and to interpret variations in the isotopic composition of riverine particulate organic carbon.
Metadata only
A numerical method for solving partial differential equations on highly irregular evolving grids
(1995-08-24) Braun, Jean; Sambridge, Malcolm
An efficient numerical method is described for solving partial differential equations in problems where traditional eulerian and lagrangian techniques fail. The approach makes use of the geometrical concept of 'natural neighbours', the properties of which make it suitable for solving problems involving large deformation and solid-fluid interactions on a deforming mesh, without the need for regridding. The approach can also be applied to high-order partial differential equations (such as the Navier-Stokes equation), even in cases where the evolving mesh is highly irregular.
Metadata only
Subspace methods for large inverse problems with multiple parameter classes
(1988) Kennett, B. L.N.; Sambridge, M. S.; Williamson, P. R.
Most nonlinear inverse problems can be cast into the form of determining the minimum of a misfit functional of model parameters. This functional determines the misfit between observations and the corresponding theoretical predictions, subject to some regularization conditions on the form of the model. When there is only one type of parameter in the model, methods based on gradient techniques work well, especially when information on rate of change of gradients is included. In the case of problems depending on multiparameter classes, simple gradient methods mix parameters of different character and physical dimensionality. This may lead to rather poor convergence and strong dependence on the scaling of the different parameter types. These difficulties can be overcome by replacing a gradient step by a local minimization in a subspace spanned by a limited number of vectors in model space. The basis vectors for the subspace should be chosen in the directions determined by the variation of the misfit functional with respect to each of the parameter types, with supplementation if required by additional vectors representing the rate of change of the gradient partitions. The construction of the perturbation requires the inversion of a matrix with the dimensions of the subspace which is easily accomplished. Such a subspace scheme takes into account the different functional dependences on the various parameter types in a balanced way. The update to the current model does not depend on the scaling of the individual parameter classes. The subspace method is flexible and can be adapted to a wide range of choices of misfit criterion and modes of representation of the parameter classes. This style of iterative subspace procedure is well adapted to nonlinear problems with dependence on many parameters and can be successfully applied in a variety of problems, e.g. seismic reflection tomography, the simultaneous nonlinear determination of earthquake locations and velocity fields and in the inversion of full seismic waveforms.
Metadata only
Genetic algorithms
(1994-01-18) Gallagher, Kerry; Sambridge, Malcolm
Genetic algorithms represent an efficient global method for nonlinear optimization problems, that are encountered in the earth sciences. They share the favorable characteristics of random Monte Carlo over local optimization methods in that they do not require linearizing assumptions nor the calculation of partial derivatives, are independent of the misfit criterion, and avoid numerical instabilities associated with matrix inversion. The additional advantages over conventional methods such as iterative least squares is that the sampling is global, rather than local, thereby reducing the tendency to become entrapped in local minima and avoiding a dependency on an assumed starting model. In contrast to random Monte Carlo, however, they also share a desirable characteristic of the local methods in that they assimilate and take advantage of information collected during the sampling of the model space, resulting in an extremely efficient and robust optimization technique. This paper describes the basic genetic algorithm, briefly highlights some recent applications in the earth sciences and concludes that, in this field, the methodology should have many applications.
Metadata only
Fire regimes of Australia
(2013-05-18) Murphy, Brett P.; Bradstock, Ross A.; Boer, Matthias M.; Carter, John; Cary, Geoffrey J.; Cochrane, Mark A.; Fensham, Roderick J.; Russell-Smith, Jeremy; Williamson, Grant J.; Bowman, David M.J.S.
Aim Comparative analyses of fire regimes at large geographical scales can potentially identify ecological and climatic controls of fire. Here we describe Australia’s broad fire regimes, and explore interrelationships and trade-offs between fire regime components. We postulate that fire regime patterns will be governed by trade-offs between moisture, productivity, fire frequency and fire intensity. Location Australia. Methods We reclassified a vegetation map of Australia, defining classes based on typical fuel and fire types. Classes were intersected with a climate classification to derive a map of ‘fire regime niches’. Using expert elicitation and a literature search, we validated each niche and characterized typical and extreme fire intensities and return intervals. Satellite-derived active fire detections were used to determine seasonal patterns of fire activity. Results Fire regime characteristics are closely related to the latitudinal gradient in summer monsoon activity. Frequent low-intensity fires occur in the monsoonal north, and infrequent, high-intensity fires in the temperate south, demonstrating a trade-off between frequency and intensity: that is, very high-intensity fires are only associated with very low-frequency fire regimes in the high biomass eucalypt forests of southern Australia. While these forests occasionally experience extremely intense fires (> 50,000 kW m-1), such regimes are exceptional, with most of the continent dominated by grass fuels, typically burning with lower intensity (< 5000 kW m-1). Main conclusions Australia’s fire regimes exhibit a coherent pattern of frequent, grass-fuelled fires in many differing vegetation types. While eucalypts are a quintessential Australian entity, their contribution as a dominant driver of high-intensity fire regimes, via their litter and bark fuels, is restricted to the forests of the continent’s southern and eastern extremities. Our analysis suggests that the foremost driver of fire regimes at the continental scale is not productivity, as postulated conceptually, but the latitudinal gradient in summer monsoon rainfall activity.
Metadata only
Reflections on four decades of land restoration in Australia
(2017) Campbell, Andrew; Alexandra, Jason; Curtis, David
The past four decades have seen a transformative process in Australian agriculture - the gradual incorporation of conservation practices such as ecological restoration, revegetation and agroforestry as a response to land degradation. Although actions have been impressive they remain fragmented, are confined to particular districts or properties and run the risk of not being built upon in the future. This paper traces the history of this movement and draws out lessons and implications for future policy development and research. Landscape-scale restoration and the integration of conservation into farming landscapes have been recognised as a global imperative for decades, for which Australia has generated many innovations - in the technical, social and policy domains. Scanning the 'big picture', we identify many pixels of best practice in policy, incentives, planning, regulation and on-ground practice. We wonder why we have not pulled these together, to work in concert over time. If we had, Australia would have a world's best natural resource management framework. However, we have neither integrated these elements at multiple scales nor sustained them. Unfortunately, although we are excellent at innovating, we have been equally good at forgetting. Progress remains partial, patchy and slow. Too often, we have made gains then gone backwards, reflecting a tendency towards policy adhockery and amnesia. With Australia's continuing depreciation of institutional memory, we risk losing critical capabilities for making sound policy decisions. Australian expertise in revegetation, restoration and regeneration of landscapes remains formidable however, with an enormous amount to offer the world. We are still learning to live and farm more sustainably, but we have made big strides over the last four decades. The challenge will be to maintain the momentum and provide adequate succession so future generations continue the work.
Metadata only
Floods after bushfires
(2020-01-02) Alexandra, Jason; Finlayson, C. Max
Metadata only
Resolving the structure of active sites on platinum catalytic nanoparticles
(2010-08-11) Chang, Lan Yun; Barnard, Amanda S.; Gontard, Lionel Cervera; Dunin-Borkowski, Rafal E.
Accurate understanding of the structure of active sites is fundamentally important in predicting catalytic properties of heterogeneous nanocatalysts. We present an accurate determination of both experimental and theoretical atomic structures of surface monatomic steps on industrial platinum nanoparticles. This comparison reveals that the edges of nanoparticles can significantly alter the atomic positions of monatomic steps in their proximity, which can lead to substantial deviations in the catalytic properties compared with the extended surfaces.
Metadata only
Institutional path dependence and environmental water recovery in Australia's Murray-Darling Basin
(2016) Marshall, Graham R.; Alexandra, Jason
The concept of institutional path dependence offers useful ways of understanding the trajectories of water policy reforms and how past institutional arrangements, policy paradigms and development patterns constrain current and future choices and limit institutional adaptability. The value of this concept is demonstrated through an analysis of environmental water recovery in Australia's Murray-Darling Basin, where while significant water volumes have been reallocated to the environment, the costs have also been significant. While there are significant lessons from the Australian experience, attempts to emulate the approach involve substantive risks and may be prohibitively costly for less wealthy nations. Context-specific institutional analysis is emphasised as fundamental to water reform and critical for reform architecture and sequencing. A key finding is that while crisis can provide powerful catalysts for institutional innovation, institutional path dependence in the absence of active and disruptive policy entrepreneurs fosters a strong tendency to reinforce the status quo and limit innovation, potentially exposing social-ecological systems to greater shocks due to climate change and other sources of escalating uncertainty.
Metadata only
Twinning, dynamic recovery and recrystallization in the hot rolling process of twin-roll cast AZ31B alloy
(2011-05-05) Zhang, Zhen; Wang, Ming Pu; Li, Zhou; Jiang, Nian; Hao, Shimeng; Gong, Jing; Hu, Hailong
Hot-rolling experiments with a reduction from 10% to 60% in single pass were conducted on AZ31B twin-roll cast sheets. Optical microscope (OM), electron backscattered diffraction (EBSD) tests and transmission electron microscope (TEM) were used to investigate twinning and DRV/DRX behaviors at different stage in the hot-rolling process. Two types of twinning occurred in the initial stage of hot-rolling process. DRV and discontinuous recrystallization dominated at moderate strain while continuous DRX took place homogeneously throughout original grains at the largest strains.
Metadata only
Reducing stress
(2020-02-01) McKimmie, Blake M.; Butler, Tamara; Chan, Edward; Rogers, Allira; Jimmieson, Nerina L.
Three studies systematically explored the relationship between social support and group identification in the context of how individuals cope with stress. In Study 1, 101 participants took part in a simulated group task where they either received social support or not under conditions of either high or low demand. Social support was associated with higher group identification, and this mediated the effect of social support on more positive appraisals and task satisfaction. In Study 2, 83 participants were either made aware of their group membership or worked as individuals on a group task under high or low demand. In this study, group membership salience was associated with greater perceived support, which was associated with greater group identification, and subsequently more positive primary and secondary appraisals, more problem-focused coping, and task satisfaction. Study 3 assessed the perceived social support and group identification reported by 71 volleyball team members before and after a game. Results were more consistent with the notion that support and identification were two concurrent perceptions associated with being in a group, rather than identification priming the recognition of support or support increasing identification.
Metadata only
A fossil flora from rafted Plio-Pleistocene mudstones at Regatta Point, Tasmania
(1985) Hill, Robert S.; Macphail, Mike
Metadata only
Large-scale comparative analysis of splicing signals and their corresponding splicing factors in eukaryotes
(2008) Schwartz, Schraga H.; Silva, Joao; Burstein, David; Pupko, Tal; Eyras, Eduardo; Ast, Gil
Introns are among the hallmarks of eukaryotic genes. Splicing of introns is directed by three main splicing signals: the 5' splice site (5'ss), the branch site (BS), and the polypyrimdine tract/3'splice site (PPT-3'ss). To Study the evolution of these splicing signals, we have conducted a systematic comparative analysis of these signals in over 1.2 million introns from 22 eukaryotes. our analyses suggest that all these signals have dramatically evolved: The PPT is weak among most fungi, intermediate in plants and protozoans, and strongest in metazoans. Within metazoans it shows a gradual strengthening from Caenorhabditis elegans to human. The 5'ss and the BS were found to be degenerate among most organisms, but highly conserved among some fungi. A maximum parsimony-based algorithm for reconstructing ancestral position-specific scoring matrices suggested that the ancestral 5'ss and BS were degenerate, as in metazoans. To shed light on the evolutionary variation in splicing signals, we have analyzed the evolutionary changes in the factors that bind these signals. Our analysis reveals coevolution of splicing signals and their corresponding splicing factors: The strength of the PPT is correlated to changes in key residues in its corresponding splicing factor U2AF2; limited correlation was found between changes in the 5'ss and Ul snRNA that binds it; but not between the BS and U2 snRNA. Thus, although the basic ability of eukaryotes to splice introns has remained conserved throughout evolution, the splicing signals and their corresponding splicing factors have considerably evolved, uniquely shaping the splicing mechanisms of different organisms.
Metadata only
Co-evolution of the branch site and SR proteins in eukaryotes
(2008) Plass, Mireya; Agirre, Eneritz; Reyes, Diana; Camara, Francisco; Eyras, Eduardo
Serine-arginine-rich (SR) proteins are essential for splicing in metazoans but are absent in yeast. By contrast, many fungi have SIR protein homologs with variable arginine-rich regions analogous to the arginine-serine-rich (RS) domain in metazoans. The density of RS repeats in these regions correlates with the conservation of the branch site signal, providing evidence for an ancestral origin of SR proteins and indicating that the SIR proteins and the branch site co-evolved.
Metadata only
Genome-Wide Association between Branch Point Properties and Alternative Splicing
(2010) Corvelo, Andre; Hallegger, Martina; Smith, Christopher W. J.; Eyras, Eduardo
The branch point (BP) is one of the three obligatory signals required for pre-mRNA splicing. In mammals, the degeneracy of the motif combined with the lack of a large set of experimentally verified BPs complicates the task of modeling it in silico, and therefore of predicting the location of natural BPs. Consequently, BPs have been disregarded in a considerable fraction of the genome-wide studies on the regulation of splicing in mammals. We present a new computational approach for mammalian BP prediction. Using sequence conservation and positional bias we obtained a set of motifs with good agreement with U2 snRNA binding stability. Using a Support Vector Machine algorithm, we created a model complemented with polypyrimidine tract features, which considerably improves the prediction accuracy over previously published methods. Applying our algorithm to human introns, we show that BP position is highly dependent on the presence of AG dinucleotides in the 3' end of introns, with distance to the 3' splice site and BP strength strongly correlating with alternative splicing. Furthermore, experimental BP mapping for five exons preceded by long AG-dinucleotide exclusion zones revealed that, for a given intron, more than one BP can be chosen throughout the course of splicing. Finally, the comparison between exons of different evolutionary ages and pseudo exons suggests a key role of the BP in the pathway of exon creation in human. Our computational and experimental analyses suggest that BP recognition is more flexible than previously assumed, and it appears highly dependent on the presence of downstream polypyrimidine tracts. The reported association between BP features and the splicing outcome suggests that this, so far disregarded but yet crucial, element buries information that can complement current acceptor site models.
Metadata only
Nonnegative realization of a linear system with nonnegative impulse response
(1996) Anderson, Brian D.O.; Deistler, Manfred; Farina, Lorenzo; Benvenuti, Luca
Let H (z) be a rational transfer function with associated nonnegative impulse response sequence. The paper considers the question: When does there exist a triple A ε R NXA b ε R N c ε R N with all nonnegative entries and H(z) = c′(zI - A) -1b? An essentially complete characterization is given of the H(z') allowing such a realization in terms of the location of the pole or poles of H(z) with maximum modulus.
Metadata only
Trace element geochemistry of CR chondrite metal
(2013) Jacquet, Emmanuel; Paulhiac-Pison, Marine; Alard, Olivier; Kearsley, Anton T.; Gounelle, Matthieu
We report trace element analyses by laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS) of metal grains from nine different CR chondrites, distinguishing grains from chondrule interior (interior grains), chondrule surficial shells (margin grains), and the matrix (isolated grains). Save for a few anomalous grains, Ni-normalized trace element patterns are similar for all three petrographic settings, with largely unfractionated refractory siderophile elements and depleted volatile Au, Cu, Ag, S. All three types of grains are interpreted to derive from a common precursor approximated by the least-melted, fine-grained objects in CR chondrites. This also excludes recondensation of metal vapor as the origin of the bulk of margin grains. The metal precursors were presumably formed by incomplete condensation, with evidence for high-temperature isolation of refractory platinum-group-element (PGE)-rich condensates before mixing with lower temperature PGE-depleted condensates. The rounded shape of the Ni-rich, interior grains shows that they were molten and that they equilibrated with silicates upon slow cooling (1-100Kh(-1)), largely by oxidation/evaporation of Fe, hence their high Pd content, for example. We propose that Ni-poorer, amoeboid margin grains, often included in the pyroxene-rich periphery common to type I chondrules, result from less intense processing of a rim accreted onto the chondrule subsequent to the melting event recorded by the interior grains. This means either that there were two separate heating events, which formed olivine/interior grains and pyroxene/margin grains, respectively, between which dust was accreted around the chondrule, or that there was a single high-temperature event, of which the chondrule margin records a late quenching phase, in which case dust accreted onto chondrules while they were molten. In the latter case, high dust concentrations in the chondrule-forming region (at least three orders of magnitude above minimum mass solar nebula models) are indicated.
Metadata only
The formation conditions of enstatite chondrites
(2015) Jacquet, Emmanuel; Alard, Olivier; Gounelle, Matthieu
We report in situ LA-ICP-MS trace element analyses of silicate phases in olivine-bearing chondrules in the Sahara 97096 (EH3) enstatite chondrite. Most olivine and enstatite present rare earth element (REE) patterns comparable to their counterparts in type I chondrules in ordinary chondrites. They thus likely share a similar igneous origin, likely under similar redox conditions. The mesostasis however frequently shows negative Eu and/or Yb (and more rarely Sm) anomalies, evidently out of equilibrium with olivine and enstatite. We suggest that this reflects crystallization of oldhamite during a sulfidation event, already inferred by others, during which the mesostasis was molten, where the complementary positive Eu and Yb anomalies exhibited by oldhamite would have possibly arisen due to a divalent state of these elements. Much of this igneous oldhamite would have been expelled from the chondrules, presumably by inertial acceleration or surface tension effects, and would have contributed to the high abundance of opaque nodules found outside them in EH chondrites. In two chondrules, olivine and enstatite exhibit negatively sloped REE patterns, which may be an extreme manifestation of a general phenomenon (possibly linked to near-liquidus partitioning) underlying the overabundance of light REE observed in most chondrule silicates relative to equilibrium predictions. The silicate phases in one of these two chondrules show complementary Eu, Yb, and Sm anomalies providing direct evidence for the postulated occurrence of the divalent state for these elements at some stage in the formation reservoir of enstatite chondrites. Our work supports the idea that the peculiarities of enstatite chondrites may not require a condensation sequence at high C/O ratios as has long been believed.
Metadata only
In situ measurement of Re-Os isotopes in mantle sulfides by laser ablation multicollector-inductively coupled plasma mass spectrometry
(2001-09-15) Pearson, NJ; Alard, O; Griffin, WL; Jackson, SE; O'Reilly, SY
A method has been developed for the in situ determination of Re-Os isotopes in single grains of sulfides in mantle-derived peridotites using a laser ablation microprobe attached to a multicollector-induced coupled plasma mass spectrometer (MC-ICPMS). High-precision Os isotope analysis by MC-ICPMS is demonstrated by the measurement of interlaboratory Os standards. Evaluation of mass bias correction procedures shows that the exponential law provides the best fit to the Os isotope data and that the ratio of the mass fractionation coefficients for Re and Os remains constant for the range of typical instrument operating conditions, This relationship enables the accurate and precise correction of the isobaric interference of Re-187 on Os-187 for Re-187/Os-188 values up to 1.6.Results are presented for single sulfide inclusions in olivine macrocryts from kimberlites in the Siberian and Slave Cratons, and sulfides enclosed in silicates and interstitial to silicates in peridotite xenoliths from the Slave Craton and Massif Central, France. Enclosed sulfides larger than 50 mum in diameter and with Os contents greater than or equal to40 ppm give Os-187/Os-188 ratios with a precision of 0.1% (2 SE), which is equivalent to N-TIMS whole-rock data. Interstitial sulfides typically have lower Os (10 to 30 ppm) and give analyses with lower precision (similar to1 to 2%) but still provide valuable information on the movement of Os within the lithosphere. The sulfide inclusions in silicates preserve significantly less radiogenic Os isotopic compositions than interstitial sulfides and accordingly produce significantly older and more realistic Re-Os age information. Whole-rock Os isotope compositions reflect the proportions of different generations of enclosed and interstitial sulfides, this calls into question the significance of many published "depletion ages." Copyright (C) 2002 Elsevier Science Ltd.

Browse

Recent Submissions