scholarly journals Comparative Document Summarisation via Classification

2019 ◽  
Vol 33 ◽  
pp. 20-28 ◽  
Author(s):  
Umanga Bista ◽  
Alexander Mathews ◽  
Minjeong Shin ◽  
Aditya Krishna Menon ◽  
Lexing Xie
Keyword(s):  
Recent Literature ◽  
Binary Classification ◽  
Subset Selection ◽  
Crowd Sourcing ◽  
Diverse Range ◽  
Discrete Optimisation ◽  
Data Subset ◽  
Gradient Based ◽  
New Formulation ◽  
Via Classification

Thispaperconsidersextractivesummarisationinacomparative setting: given two or more document groups (e.g., separated by publication time), the goal is to select a small number of documents that are representative of each group, and also maximally distinguishable from other groups. We formulate a set of new objective functions for this problem that connect recent literature on document summarisation, interpretable machine learning, and data subset selection. In particular, by casting the problem as a binary classification amongst different groups, we derive objectives based on the notion of maximum mean discrepancy, as well as a simple yet effective gradient-based optimisation strategy. Our new formulation allows scalable evaluations of comparative summarisation as a classification task, both automatically and via crowd-sourcing. To this end, we evaluate comparative summarisation methods on a newly curated collection of controversial news topics over 13months.Weobserve thatgradient-based optimisationoutperforms discrete and baseline approaches in 15 out of 24 different automatic evaluation settings. In crowd-sourced evaluations, summaries from gradient optimisation elicit 7% more accurate classification from human workers than discrete optimisation. Our result contrasts with recent literature on submodular data subset selection that favours discrete optimisation. We posit that our formulation of comparative summarisation will prove useful in a diverse range of use cases such as comparing content sources, authors, related topics, or distinct view points.

Data Subset Selection With Imperfect Multiple Labels

2019 ◽  
Vol 30 (7) ◽  
pp. 2212-2221 ◽  
Author(s):  
Meng Fang ◽  
Tianyi Zhou ◽  
Jie Yin ◽  
Yang Wang ◽  
Dacheng Tao
Keyword(s):  
Subset Selection ◽  
Data Subset

Foam and liquid sclerotherapy for varicose veins

2009 ◽  
Vol 24 (1_suppl) ◽  
pp. 62-72 ◽  
Author(s):  
P Coleridge Smith
Keyword(s):  
Medical Literature ◽  
Recent Literature ◽  
Varicose Veins ◽  
Visual Disturbance ◽  
Optimum Ratio ◽  
Ultrasound Guided ◽  
Foam Sclerotherapy ◽  
Serious Adverse Events ◽  
Diverse Range ◽  
Systemic Complications

Objective The objective of this study is to review the methods and outcome of ultrasound-guided foam sclerotherapy (UGFS) for the treatment of superficial venous incompetence. Method Medical literature databases including Medline were searched for recent literature concerning UGFS. Papers describing methods and outcome have been assessed and their main findings included in this summary. A detailed description of the methods used by the author has been included as an example of how successful the treatment may be achieved. Results A diverse range of practice is described in published literature in this field. Each group of authors used their own variation of the methods, described in the published literature, with good results. It is clear that foam sclerotherapy is far more effective than liquid sclerotherapy and that ultrasound imaging allows the treatment to be delivered accurately to affected veins. There is evidence that 3% policocanol foam is no more effective than 1% polidocanol foam. The optimum ratio of gas to liquid is 4:1, although a range of ratios is reported in published work. There is a wide variation in the volume used as well as the method by which it is injected. The use of carbon dioxide foam reduces the systemic complications, particularly visual disturbance, when compared with air foams. Very few serious adverse events have been reported in the literature despite the widespread use of this method. Rates of recanalization of saphenous trunks following UGFS are similar to those observed after endovenous laser and endovenous radiofrequency ablation of veins, as well as the residual incompetence after surgical treatment. Conclusions UGFS is a safe and effective method of treating varicose veins. The relative advantages or disadvantages of this treatment in the longer term are yet to be published.

scholarly journals Multi-level gradient-based methods and parametrisation in aerodynamic shape design

2008 ◽  
pp. 169-197
Author(s):  
Massimiliano Martinelli ◽  
François Beux
Keyword(s):  
Shape Design ◽  
Inviscid Flows ◽  
Shape Representations ◽  
Aerodynamic Shape ◽  
Gradient Based ◽  
Multi Level ◽  
Grid Points ◽  
New Formulation ◽  
Adequate Sets ◽  
Compact Shape

The present study focuses on multi-level approaches in the context of discrete gradient-based methods for aerodynamic shape design. More precisely, the minimisation is done alternatively on different control subspaces according to multigrid-like cycles, providing at each sub-level a particular gradient preconditioning. Starting from an existing multi-level gradient-based formulation associated to shape grid-points coordinates, a possible generalisation to more compact shape representations is proposed through the construction of adequate sets of embedded shape sub-parametrisations. The behaviour of the new formulation is illustrated on different 2D inverse problems for inviscid flows.

scholarly journals A Submodularity Framework for Data Subset Selection

2013 ◽  
Author(s):  
Katrin Kirchhoff ◽  
Jeff Bilmes ◽  
Kai Wei ◽  
Yuzong Liu ◽  
Arindam Mandal ◽  
...  
Keyword(s):  
Subset Selection ◽  
Data Subset

Topology Optimization of Micromachined Structures With Surface Micromachining Manufacturing Constraints

2006 ◽  
Author(s):  
Aravind Alwan ◽  
G. K. Ananthasuresh
Keyword(s):  
Topology Optimization ◽  
Optimization Methods ◽  
Manufacturing Cost ◽  
Manufacturing Constraints ◽  
Surface Micromachining ◽  
Out Of Plane ◽  
Gradient Based ◽  
Plane Direction ◽  
New Formulation ◽  
Design Parameterization

Topology optimization methods have been shown to have extensive application in the design of microsystems. However, their utility in practical situations is restricted to predominantly planar configurations due to the limitations of most microfabrication techniques in realizing structures with arbitrary topologies in the direction perpendicular to the substrate. This study addresses the problem of synthesizing optimal topologies in the out-of-plane direction while obeying the constraints imposed by surface micromachining. A new formulation that achieves this by defining a design space that implicitly obeys the manufacturing constraints with a continuous design parameterization is presented in this paper. This is in contrast to including manufacturing cost in the objective function or constraints. The resulting solutions of the new formulation obtained with gradient-based optimization directly provide the photolithographic mask layouts. Two examples that illustrate the approach for the case of stiff structures are included.

scholarly journals Crowd-sourcing structure-from- motion data for terrain modelling in a real-world disaster scenario: A proof of concept

2019 ◽  
Vol 43 (2) ◽  
pp. 236-259
Author(s):  
JJ Ratner ◽  
JJ Sury ◽  
MR James ◽  
TA Mather ◽  
DM Pyle
Keyword(s):  
Risk Reduction ◽  
Structure From Motion ◽  
Real World ◽  
Disaster Risk Reduction ◽  
High Standard ◽  
Point Clouds ◽  
Disaster Risk ◽  
Crowd Sourcing ◽  
Diverse Range ◽  
Terrain Attributes

Structure-from-motion (SfM) photogrammetry techniques are now widely available to generate digital terrain models (DTMs) from optical imagery, providing an alternative to costlier options such as LiDAR or satellite surveys. SfM could be a useful tool in hazard studies because its minimal cost makes it accessible even in developing regions and its speed of use can provide updated data rapidly in hazard-prone regions. Our study is designed to assess whether crowd-sourced SfM data is comparable to an industry standard LiDAR dataset, demonstrating potential real-world use of SfM if employed for disaster risk reduction purposes. Three groups with variable SfM knowledge utilized 16 different camera models, including four camera phones, to collect 1001 total photos in one hour of data collection. Datasets collected by each group were processed using VisualSFM, and the point densities, accuracies and distributions of points in the resultant point clouds (DTM skeletons) were compared. Our results show that the point clouds are resilient to inconsistency in users’ SfM knowledge: crowd-sourced data collected by a moderately informed general public yields topography results comparable in data density and accuracy to those produced with data collected by highly-informed SfM users or experts using LiDAR. This means that in a real-world scenario involving participants with a diverse range of expertise, topography models could be produced from crowd-sourced data quite rapidly and to a very high standard. This could be beneficial to disaster risk reduction as a relatively quick, simple and low-cost method to attain rapidly updated knowledge of terrain attributes, useful for the prediction and mitigation of many natural hazards.

Sign in / Sign up

Export Citation Format

Share Document