58622 research outputs found
Sort by
Fast Inexact Bilevel Optimization for Analytical Deep Image Priors
Get PDFThe analytical deep image prior (ADP) introduced by Dittmer et al. (2020) establishes a link between deep image priors and classical regularization theory via bilevel optimization. While this is an elegant construction, it involves expensive computations if the lower-level problem is to be solved accurately. To overcome this issue, we propose to use adaptive inexact bilevel optimization to solve ADP problems. We discuss an extension of a recent inexact bilevel method called the method of adaptive inexact descent of Salehi et al.(2024) to an infinite-dimensional setting required by the ADP framework. In our numerical experiments we demonstrate that the computational speed-up achieved by adaptive inexact bilevel optimization allows one to use ADP on larger-scale problems than in the previous literature, e.g. in deblurring of 2D color images
Constructed wetlands as a nature-based treatment solution for emerging contaminants in wastewater: chemical fate and ecotoxicological impacts
No full textMethods and tools for analysis and visualisation of oceanographic and acoustic measurements from the HiAOOS mooring network in the Arctic Ocean
Get PDFThe EU-funded project High Arctic Ocean Observation System (HiAOOS) aims to implement a large-scale network in the Nansen and Amundsen Basin, including four multidisciplinary moorings in the deep basins and two experimental moorings north of Svalbard. Oceanographic sensors provide measurements of temperature, salinity, dissolved oxygen, ocean currents, and sea ice while acoustic instruments measuring ocean sound facilitate for acoustic thermometry, underwater geopositioning, detection of underwater earthquakes, and monitoring of marine mammals and human activities.Based on data collections from the earlier projects (CAATEX, INTAROS) and new observations from HiAOOS, a variety of methods and tools is developed for analysis of continuous, year-round mooring measurements. New tools for data processing and integrative analysis of ocean and sea ice measurements allow for ingestion of auxiliary products to explore data, generate statistics, and provide advanced visualisations. New methods and tools are developed for detection of earthquakes signatures and automatic identification of sound sources in passive acoustic recordings to understand the environmental impacts and contribution from human activities. New methods for inversions of long-range acoustic receptions to ocean temperature are developed for ocean thermometry. New techniques for detection of low frequency acoustic signals are employed to test ranging and localisation for underwater geopositioning. Methodology for validation of ocean-ice reanalysis using acoustic receptions is established. Selected data and methods are ingested into Blue Insight, a digital modular platform for processing, visualization and sharing of ocean information. The poster will present the overview of methods and tools developed in HiAOOS for enhanced use of new observations.<br/
Dataset for "Covering one point process with another"
No full textFix a set A as a subset of Euclidean space (for example, a disc or a polygon), and a subset B contained in A (we often consider the case B=A). Place n points X_1, ..., X_n in A, with the locations chosen independently and uniformly at random. We think of these as "transmitters". Place another m points Y_1, ..., Y_m in B contained in A, which we think of as "receivers". At each X_i, place a Euclidean ball of radius r. We define R_{n,m,k} to be the smallest r such that every receiver Y_j has at least k transmitters within distance r. In the paper "Covering one point process with another" we proved that if m/n tends to tau as n tends to infinity, then the quantity n R_{n,m,k}^d - c_1 log(n) - c_2 loglog(n) (for constants c_1,c_2 which we give in the paper) converges to a random variable (whose distribution we also give). These datasets include large numbers of independent samples of n R_{n,m,k}^d - c_1 log(n) - c_2 loglog(n). The dataset is separated into files, and each file into rows. All the data in a given file are generated using fixed sets A and B, and parameters n, m, d, k. Each row in this given file is a single number: the outcome of an experiment, conducted independently of the other rows. In each experiment we place n points at random locations in A, place m points at random locations in B, calculate R_{n,m,k} as described above (and as detailed formally in the paper) and record the value of n R_{n,m,k}^d - c_1 log(n) - c_2 loglog(n) on a row. For the next row, we remove the existing points, and place n points in A, m points in B, etc. for the same n,m, A, B, but with the random points chosen independently of previous experiments. In probabilisitic terms, the rows of a given file are independent and identically distributed random variables with a common distribution, which is the distribution of n R_{n,m,k}^d - c_1 log(n) - c_2 loglog(n). The distribution depends on A, B, n, m, d and k. Different files were generated using different choices of A, B, n, m, d and k. The paper was written by Frankie Higgs, Mathew D. Penrose and Xiaochuan Yang. We thank Keith Briggs for suggesting the problem and advice on the simulations
Country-branded universities:a framework for country brand authenticity in transnational higher education
Get PDFCountry-branded universities, which include a country in their name, are a type of higher education institution found in many countries globally. We consider these universities as transnational higher education institutions, for they ‘borrow’ or ‘transfer’ elements of the affiliated country across national borders. Surprisingly, there is very limited research on this phenomenon. The aims of this research are to identify what exactly in practice crosses national borders from the affiliated country to a country-branded university, and to develop a framework for country-brand authenticity in transnational higher education (TNHE). The data used in this research was provided by expert informants, who were individuals holding a senior or middle management position at a country-branded university. The proposed framework for country-brand authenticity in TNHE has four main dimensions, namely affiliated country connections; teaching and learning; governance and human resources; and student experience and development. These four dimensions are supported by eleven themes. The most often mentioned concepts were related to the curricula, pedagogy/didactics, and foreign accreditation themes. We offer suggestions for further research using our country-brand authenticity framework. <br/
Access Repository Dashboard
Do you manage Open Research Online? Become a CORE Member to access insider analytics, issue reports and manage access to outputs from your repository in the CORE Repository Dashboard! 👇