Given the challenge of gathering labelled training data for machine learning tasks, active learning has become popular. This paper focuses on the beginning of unsupervised active learning, where ...there are no labelled data at all. The aim of this zero initialised unsupervised active learning is to select the most informative examples - even from an imbalanced dataset - to be labelled manually. Our solution with proposed selection strategy, called Optimally Balanced Entropy-Based Sampling (OBEBS) reaches a balanced training set at each step to avoid imbalanced problems. Two theorems of the optimal solution for selection strategy are also presented and proved in the paper. At the beginning of the active learning, there is not enough information for supervised machine learning method, thus our selection strategy is based on unsupervised learning (clustering). The cluster membership likelihoods of the items are essential for the algorithm to connect the clusters and the classes, i.e., to find assignment between them. For the best assignment, the Hungarian algorithm is used, and single, multi, and adaptive assignment variants of OBEBS method are developed. Based on generated and real images datasets of handwritten digits, the experimental results show that our method surpasses the state-of-the-art methods.
Forests provide multiple services, and in the face of global change adaptive management strategies are needed, which inevitably must be based on models. However, most locally accurate forest models ...are tied to the stand scale and cannot readily be applied across large areas. Empirical data for model initialisation are often not available at large spatial scales. National Forest Inventories (NFIs) provide spatially representative tree and stand samples, but their samples are typically small, that is, only a few trees are measured per plot, and they are truncated, that is, not each tree has the same probability of being observed. To overcome these issues, we develop and apply a methodology to derive stand descriptions from small sample data, taking the Swiss NFI as a case study.
We extended the traditional Weibull function to (multi‐)truncated unimodal and bimodal forms that are suitable for the representation of samples from survey designs with multiple callipering thresholds. Subsequently, we applied these functions in an extended parameter prediction method to derive stand diameter distributions from representative samples. Additionally, we predicted species compositions using a multinomial logistic regression model and assigned them to the diameter distributions of the stands.
The diameter distribution of 9.1% of the Swiss NFI samples was better described by a bimodal than a unimodal Weibull function. The uni‐ and bimodal diameter model in combination with the model to determine species composition can be used to predict stand descriptions from single small samples or entire forest types in the target area. Thereby, the bimodal form is suitable for capturing stand structures with distinct under‐ and overstorey. In Switzerland, the diameter distributions of stands are typically positively skewed.
Our method can be applied to any large‐scale dataset (e.g. NFI) and allows to generate initial conditions in terms of spatially representative stands. These, in turn, are suitable for forest stand simulators, which allows for developing adaptive forest management strategies at large scales, by simulating realistic and site‐specific stand development while still reflecting detailed management measures. Furthermore, stand descriptions can be used to assess tree species diversity, regeneration and harvest potentials.
Zusammenfassung
Wälder erbringen vielfältige Leistungen und im Angesicht globaler Veränderungen werden adaptive Bewirtschaftungsstrategien gebraucht, die zwangsläufig auf Modellen beruhen müssen. In der Regel sind lokal präzise Waldmodelle jedoch an die Bestandesebene gebunden und lassen sich nicht direkt auf großräumige Regionen übertragen. Empirische Daten zur Modellinitialisierung sind auf großen räumlichen Skalen oft nicht verfügbar. Nationale Waldinventuren (NFIs) liefern räumlich repräsentative Baum‐ und Bestandesstichproben, diese sind aber typischerweise klein (nur wenige Bäume pro Aufnahmefläche) und trunkiert (nicht jeder Baum hat dieselbe Aufnahmewahrscheinlichkeit). Um diese Lücke zu füllen, entwickeln wir eine Methodik zur Herleitung von Bestandesbeschreibungen aus kleinen Stichprobendaten und zeigen am Beispiel des Schweizerischen Landesforestinventars (NFI) eine konkrete Anwendung.
Wir erweiterten die Weibull‐Funktion zu einer (multi‐)trunkierten unimodalen und bimodalen Form, die sich für die Beschreibung von Stichproben aus Aufnahmedesigns mit mehreren Kluppschwellen eignen. Anschließend wendeten wir diese Funktionen in einer simultanen Parameter‐Schätzmethode über alle Stichproben an, um repräsentative Bestandes‐Durchmesserverteilungen abzuleiten. Zusätzlich prognostizierten wir die Baumartenzusammensetzung mit Hilfe eines multinomialen logistischen Regressionsmodells und verbanden diese mit den Durchmesserverteilungen der Bestände.
Die Durchmesserverteilung von 9,1% der Schweizerischen NFI Stichproben wurde besser durch eine bimodale als durch eine unimodale Weibull‐Funktion beschrieben. Das uni‐ und bimodale Durchmessermodell in Kombination mit dem Modell zur Bestimmung der Baumartenzusammensetzung kann zur Vorhersage von Bestandesbeschreibungen aus einzelnen kleinen Stichproben oder ganzen Waldtypen im Zielgebiet verwendet werden. Dabei eignet sich die bimodale Form zur Erfassung von Bestandesstrukturen mit ausgeprägtem Unter‐ und Oberstand. In der Schweiz sind die Durchmesserverteilungen von Beständen typischerweise rechtsschief.
Unsere Methode kann auf jeglichen grossräumigen Datensatz (z.B. NFI) angewendet werden und erlaubt es, Initialbedingungen in Form von räumlich repräsentativen Beständen zu generieren. Diese wiederum können in Waldbestandessimulatoren verwendet werden, um die Entwicklung von adaptiven Waldbewirtschaftungsstrategien auf großräumigen Skalen zu untersuchen, indem realistische und standortspezifische Bestandesentwicklungen mit zielgenauer Bewirtschaftung simuliert werden. Darüber hinaus können die Bestandesbeschreibungen zur Beurteilung der Baumartenvielfalt, der Verjüngung und des Erntepotentials genutzt werden.
The exact signalling mechanism of the mTOR complex remains a subject of constant debate, even with some evidence that amino acids participate in the same pathway as used for insulin signalling during ...protein synthesis. Therefore, this work conducted further study of the actions of amino acids, especially leucine, in vivo, in an experimental model of cachexia. We analysed the effects of a leucine-rich diet on the signalling pathway of protein synthesis in muscle during a tumour growth time-course.
Wistar rats were distributed into groups based on Walker-256 tumour implant and subjected to a leucine-rich diet and euthanised at three different time points following tumour development (the 7th, 14th and 21st day). We assessed the mTOR pathway key-proteins in gastrocnemius muscle, such as RAG-A-GTPase, ERK/MAP4K3, PKB/Akt, mTOR, p70S6K1, Jnk, IRS-1, STAT3, and STAT6 comparing among the experimental groups. Serum WF (proteolysis-induced factor like from Walker-256 tumour) and muscle protein synthesis and degradation were assessed.
The tumour-bearing group had increased serum WF content, and the skeletal-muscle showed a reduction in IRS-1 and RAG activation, increased PKB/Akt and Erk/MAP4K3 on the 21st day, and maintenance of p70S6K1, associated with increases in muscle STAT-3 and STAT-6 levels in these tumour-bearing rats.
Meanwhile, the leucine-rich diet modulated key steps of the mTOR pathway by triggering the increased activation of RAG and mTOR and maintaining JNK, STAT-3 and STAT-6 levels in muscle, leading to an increased muscle protein synthesis, reducing the degradation during tumour evolution in a host, minimising the cancer-induced damages in the cachectic state.
Efficient generation of jamming signal is an important but intractable issue in active deception jamming against synthetic aperture radar. Considerations must be given to both the computational ...complexity and the focus depth of the false scatterers of a deception template. However, existing methods cannot meet both demands mentioned above when generating jamming signal of an extended false scene or scattered false targets. In this study, a frequency-domain three-stage algorithm (FDTSA) is proposed. In theory, the jammer system is deliberately reformatted in the two-dimensional frequency domain. Accordingly, the implementation of the FDTSA can be effectively accelerated by fast Fourier transform and by separating the modulation process of a repeat jammer into three stages: the offline stage, the initialisation stage and the real-time modulation stage. Theoretical analyses and simulation results indicate that the FDTSA can get rid of severe focus deterioration of the false scatterers and reasonable computational load is required.
Purpose.
Augmented Reality (AR) in Laparoscopic Liver Resection requires anatomical landmarks and the silhouette to be found on the laparoscopic image. They are used to register the preoperative 3D ...model obtained from CT segmentation. The existing AR systems rely on the surgeon to 1) annotate the landmarks and silhouette and 2) provide an initial registration. These non-trivial tasks require surgeon attention which may perturb the procedure. We propose methods to solve both tasks, hence registration, automatically.
Methods.
The landmarks are the lower ridge and the falciform ligament. We solve 1) by training a U-Net from a new dataset of 1415 labelled images extracted from 68 procedures. We solve 2) by a novel automatic coarse-to-fine pose estimation method, including visibility-reasoning within an iterative robust process. In addition, we propose to divide the ridge into six anatomical sub-parts, making its annotation and use in registration more accurate.
Results.
Our method detects the silhouette with an error equivalent to an experienced surgeon. It detects the ridge and ligament with higher errors owing to under-detection. Nonetheless, our method successfully initialises the registration with tumour target registration errors of 22.4, 14.8 and 7.2 mm for 3 clinical procedures. In comparison, the errors from manual initialisation are 30.5, 15.1 and 16.3 mm.
Conclusion.
Our results are promising, suggesting that we have found an appropriate methodological approach.
While interior-point methods share the same fundamentals, the implementation determines the actual performance. In order to attain the highest efficiency, different applications may require ...differently tuned implementations. In this paper we describe an implementation specifically designed for
optimisation in radiation therapy
. These problems are large-scale nonlinear (and sometimes nonconvex) constrained optimisation problems, consisting of both sparse and dense data. Several application-specific properties are exploited to enhance efficiency. Permuting, tiling and mixed precision arithmetic allow the algorithm to optimally process the mixed dense and sparse data matrices (making this step 2.2 times faster, and overall runtime reduction of
55
%
) and scalability (16 threads resulted in a speed-up factor of 9.8 compared to singlethreaded performance, against a speed-up factor of 7.7 for the less optimised implementation). Predefined cost-functions are hard-coded and the computationally expensive second derivatives are written in canonical form, and combined if multiple cost-functions are defined for the same clinical structure. The derivatives are then computed using a scaled matrix–matrix product. A cheap initialisation strategy based on the background knowledge reduces the number of iterations by
11
%
. We also propose a novel combined Mehrotra–Gondzio approach. The algorithm is extensively tested on a dataset consisting of 120 patients, distributed over 6 tumour sites/approaches. This test dataset is made publicly available.
Plasma electrolytic polishing (PeP) is mainly used to improve the surface quality and thus the performance of electrically conductive parts. It is usually used as an anodic process, i.e., the ...workpiece is positively charged. However, the process is susceptible to high current peaks during the formation of the vapour–gaseous envelope, especially when polishing workpieces with a large surface area. In this study, the influence of the anode immersion speed on the current peaks and the average power during the initialisation of the PeP process is investigated for an anode the size of a microreactor mould insert. Through systematic experimentation and analysis, this work provides insights into the control of the initialisation process by modulating the anode immersion speed. The results clarify the relationship between immersion speed, peak current, and average power and provide a novel approach to improve process efficiency in PeP. The highest peak current and average power occur when the electrolyte splashes over the top of the anode and not, as expected, when the anode touches the electrolyte. By immersion of the anode while the voltage is applied to the anode and counterelectrode, the reduction of both parameters is over 80%.
Operational forecasting centres are currently developing data assimilation systems for coupled atmosphere-ocean models. Strongly coupled assimilation, in which a single assimilation system is applied ...to a coupled model, presents significant technical and scientific challenges. Hence weakly coupled assimilation systems are being developed as a first step, in which the coupled model is used to compare the current state estimate with observations, but corrections to the atmosphere and ocean initial conditions are then calculated independently. In this paper, we provide a comprehensive description of the different coupled assimilation methodologies in the context of four-dimensional variational assimilation (4D-Var) and use an idealised framework to assess the expected benefits of moving towards coupled data assimilation. We implement an incremental 4D-Var system within an idealised single-column atmosphere-ocean model. The system has the capability to run both strongly and weakly coupled assimilations as well as uncoupled atmosphere- or ocean-only assimilations, thus allowing a systematic comparison of the different strategies for treating the coupled data assimilation problem. We present results from a series of identical twin experiments devised to investigate the behaviour and sensitivities of the different approaches. Overall, our study demonstrates the potential benefits that may be expected from coupled data assimilation. When compared to uncoupled initialisation, coupled assimilation is able to produce more balanced initial analysis fields, thus reducing initialisation shock and its impact on the subsequent forecast. Single observation experiments demonstrate how coupled assimilation systems are able to pass information between the atmosphere and ocean and therefore use near-surface data to greater effect. We show that much of this benefit may also be gained from a weakly coupled assimilation system, but that this can be sensitive to the parameters used in the assimilation.
Obtaining a fit-for-purpose rock-type classification that adequately incorporates the key depositional and diagenetic heterogeneities is a prime challenge for carbonate reservoirs. Another prevailing ...issue is to integrate the static and dynamic data consistently with the rock-typing scheme in order to correctly initialise the reservoir flow simulation model. This paper describes a novel near-wellbore rock-typing and upscaling approach adopted to address the crucial challenges of integrating reservoir rock-typing and simulation in carbonate reservoirs. We demonstrate this workflow through a case study for a highly heterogeneous Eocene-Oligocene limestone reservoir, Field X. Geological studies carried out in Field X suggested that the key permeability pathways are strongly related to the mechanism of reservoir porosity and permeability evolution during late-burial corrosion. The rock-typing and upscaling methodology described in this paper involves the geological-petrophysical classification of the key reservoir heterogeneities through systematic evaluation of the main paragenetic events. Associations between the depositional and late-burial corrosion features, and their impact on reservoir flow properties, were accounted for in our workflow. Employing near-wellbore rock-typing and upscaling workflow yielded consistent initialisation of the Field X reservoir simulation model and therefore improved the accuracy of fluids-in-place calculation. Subsequently, the cumulative production curves computed by the reservoir simulation model of Field X showed closer agreement to the historic production data. The revised Field X simulation model is now much better constrained to the reservoir geology and provides an improved geological-prior for history matching.
•Novel rock-typing workflow integrated with near-wellbore upscaling.•Workflow successfully applied to giant carbonate field with long production history.•Improved accuracy of FIP calculations during reservoir simulation.•Improved geological-prior obtained for history matching and forecasting studies.
The impact of realistic atmospheric initialisation on the seasonal prediction of tropical Pacific sea surface temperatures is explored with the Predictive Ocean-Atmosphere Model for Australia (POAMA) ...dynamical seasonal forecast system. Previous versions of POAMA used data from an Atmospheric Model Intercomparison Project (AMIP)-style simulation to initialise the atmosphere for the hindcast simulations. The initial conditions for the hindcasts did not, therefore, capture the true intra-seasonal atmospheric state. The most recent version of POAMA has a new Atmosphere and Land Initialisation scheme (ALI), which captures the observed intra-seasonal atmospheric state. We present the ALI scheme and then compare the forecast skill of two hindcast datasets, one with AMIP-type initialisation and one with realistic initial conditions from ALI, focussing on the prediction of El Niño. For eastern Pacific (Niño3) sea surface temperature anomalies (SSTAs), both experiments beat persistence and have useful SSTA prediction skill (anomaly correlations above 0.6) at all lead times (forecasts are 9 months duration). However, the experiment with realistic atmospheric initial conditions from ALI is an improvement over the AMIP-type initialisation experiment out to about 6 months lead time. The improvements in skill are related to improved initial atmospheric anomalies rather than an improved initial mean state (the forecast drift is worse in the ALI hindcast dataset). Since we are dealing with a coupled system, initial atmospheric errors (or differences between experiments) are amplified though coupled processes which can then lead to long lasting errors (or differences).
PDF
