2,214 Works

Improving argument overlap for proposition-based summarisation

Y Fang & Simone Teufel
We present improvements to our incremental proposition-based summariser, which is inspired by Kintsch and van Dijk's (1978) text comprehension model. Argument overlap is a central concept in this summariser. Our new model replaces the old overlap method based on distributional similarity with one based on lexical chains. We evaluate on a new corpus of 124 summaries of educational texts, and show that our new system outperforms the old method and several stateof-the-art non-proposition-based summarisers. The...

Response of a flat plate wing to a transverse gust at low reynolds numbers

Simon Corkery, Holger Babinsky & J Harvey
© 2018, American Institute of Aeronautics and Astronautics Inc, AIAA. All rights reserved. In this article the unsteady response of a flat plate wing encountering a transverse gust is presented. The aim of this study is to understand how the lift and drag forces of a wing respond to a large amplitude wind gust of magnitude equivalent to the flight speed and at Reynolds numbers relevant to biological fliers and unmanned aerial vehicles. A gust...

Domain Adaptive Inference for Neural Machine Translation

Danielle Saunders, Felix Stahlberg, Adria De Gispert & Bill Byrne
We investigate adaptive ensemble weighting for Neural Machine Translation, addressing the case of improving performance on a new and potentially unknown domain without sacrificing performance on the original domain. We adapt sequentially across two Spanish-English and three English-German tasks, comparing unregularized fine-tuning, L2 and Elastic Weight Consolidation. We then report a novel scheme for adaptive NMT ensemble decoding by extending Bayesian Interpolation with source information, and show strong improvements across test domains without access to...

Automatic Discovery of the Statistical Types of Variables in a Dataset.

Isabel Valera & Zoubin Ghahramani
A common practice in statistics and machine learning is to assume that the statistical data types (e.g., ordinal, categorical or real-valued) of variables, and usually also the likelihood model, is known. However, as the availability of real- world data increases, this assumption becomes too restrictive. Data are often heterogeneous, complex, and improperly or incompletely documented. Surprisingly, despite their practical importance, there is still a lack of tools to automatically discover the statistical types of, as...

A Birth-Death Process for Feature Allocation.

Konstantina Palla, David A Knowles & Zoubin Ghahramani
We propose a Bayesian nonparametric prior over feature allocations for sequential data, the birth- death feature allocation process (BDFP). The BDFP models the evolution of the feature allocation of a set of N objects across a covariate (e.g. time) by creating and deleting features. A BDFP is exchangeable, projective, stationary and reversible, and its equilibrium distribution is given by the Indian buffet process (IBP). We show that the Beta process on an extended space is...

Spatio-Temporal Deep Graph Infomax

Felix L Opolka, Aaron Solomon, Cătălina Cangea, Petar Veličković, Pietro Liò & R Devon Hjelm
Spatio-temporal graphs such as traffic networks or gene regulatory systems present challenges for the existing deep learning methods due to the complexity of structural changes over time. To address these issues, we introduce Spatio-Temporal Deep Graph Infomax (STDGI)---a fully unsupervised node representation learning approach based on mutual information maximization that exploits both the temporal and spatial dynamics of the graph. Our model tackles the challenging task of node-level regression by training embeddings to maximize the...

Deep Graph Infomax

Petar Veličković, William Fedus, William L Hamilton, Pietro Liò, Yoshua Bengio & R Devon Hjelm
We present Deep Graph Infomax (DGI), a general approach for learning node representations within graph-structured data in an unsupervised manner. DGI relies on maximizing mutual information between patch representations and corresponding high-level summaries of graphs---both derived using established graph convolutional network architectures. The learnt patch representations summarize subgraphs centered around nodes of interest, and can thus be reused for downstream node-wise learning tasks. In contrast to most prior approaches to unsupervised learning with GCNs, DGI...

Research Support Handy Guides: How to Spot Predatory Publishers

Claire Sewell
Accessible guide to spotting predatory publishers from the Office of Scholarly Communication, Cambridge University Libraries.

Bayesian Hybrid Matrix Factorisation for Data Integration

Thomas Brouwer & Pietro Lio
We introduce a novel Bayesian hybrid matrix factorisation model (HMF) for data integration, based on combining multiple matrix factorisation methods, that can be used for in- and out-of-matrix prediction of missing values. The model is very general and can be used to integrate many datasets across different entity types, including repeated experiments, similarity matrices, and very sparse datasets. We apply our method on two biological applications, and extensively compare it to state-of-the-art machine learning and...

BIM as an Enabler for Digital Transformation

James Heaton, Ajith Parlikad, David Owens & Neil Pawsey
Organisations all over the world are increasingly becoming digitally enabled, including infrastructure providers and are looking to use this new found a digital way of working to transform the organisation into a more lean, efficient and productive organisation. Digital transformation is not exclusively about digital technology but the fact that technology, which is digital, will enable the organisation to create greater informed decisions around there current and future challenges, objectives and strategy. While many organisations...

A Multi Agent System architecture to implement Collaborative Learning for social industrial assets

K Bakliwal, MH Dhada, AS Palau, Ajith Parlikad & BK Lad
The `Industrial Internet of Things' aims to connect industrial assets with one another and subsequently bene t from the data that is generated, and shared, among these assets. In recent years, the extensive instrumentation of machines and the advancements in Information Communication Technologies are re-shaping the role of assets in our industrial systems. An emerging paradigm here is the concept of `social assets': assets that collaborate with each other in order to improve system performance....

Top-down Practical Methodology for the Development of Asset Information Requirements

James Heaton & Ajith Parlikad
During the last decade Building Information Modelling (BIM) has emerged as a key enabler within the construction industry to increase productivity. While the benefits have been mostly realised within the Design/Construction phase, the adoption of BIM within the operational phase has been significantly limited. One of the critical challenges to utilising BIM within the operational phase is for asset own-ers/maintainer to develop information requirements that support the BIM process. UK BIM standard PAS 1192-3 provides...

Dynamic maintenance based on criticality in electricity networks

J Adams & Ajith Parlikad
The need to prioritize maintenance activities and investments based on asset criticality and associated risk is seen increasingly as important in industry. However, proper use of criticality in developing maintenance strategies and plans is still at a nascent stage in most organisations. A review of industrial practices showed that criticality is considered as more or less a static quantity that is not updated with sufficient frequency as the operating environment changes. This paper examines an...

Heuristic optimisation for multi-asset intervention planning in a petrochemical plant

Sanyapong Petchrompo & Ajith Parlikad
Large infrastructure assets commonly require high intervention costs, but the absence of an effective asset management plan can bring about a massive production loss for a company. Hence, managing these assets is considered a daunting task and is even more complicated if these assets operate collectively to produce an output. This paper explores a pragmatic approach to a multi-asset intervention scheduling problem through a case study of a vessel fleet in a petrochemical plant. After...

Structuring Data for Intelligent Predictive Maintenance in Asset Management

OO Aremu, AS Palau, Ajith Parlikad, D Hyland-Wood & McAree, PR
Predictive maintenance (PdM) within asset management improves savings in operational cost, productivity, and safety management capabilities. While PdM can be administered using various methods, growing interest in Arti cial Intelligence (AI) has lead to current state of the art PdM relying on machine learning (ML) technology. Like other tools used in PdM for asset management, standards for applying ML technology for PdM are required. This work introduces a standard of practice in regards to usage...

Towards Dynamic Criticality-Based Maintenance Strategy for Industrial Assets

Joel Adams, Rengarajan Srinivasan, Ajith Parlikad, V Diaz & A Crespo Marquez
An asset’s risk is a useful indicator for determining optimal time of repair/replacement for assets in order to yield minimal operational cost of maintenance. For a successful asset management practice, asset-intensive organisations must understand the risk profile associated with their asset portfolio and how this will change over time. Unfortunately, in many risk-based asset management approaches, the only thing that is known to change in the risk profile of the asset is the likelihood (or...

Scoping study into Deriving Transport Benefits from Big Data and the Internet of Things in Smart Cities

N Hill, G Gibson, E Guidorzi, S Amaral, Ajith Parlikad & Ying Jin
Department for Transport

A bibliographic review of trends in the application of ‘criticality’ towards the management of engineered assets

Joel Adams, Ajith Parlikad & J Amadi-Echendu
Increasing budgetary constraints have raised the hiatus for allocation of funding and prioritisation of investments to ensure that long established and new assets are in the condition to provide uninterrupted services towards progressive economic and social activities. Whereas a key challenge remains how to allocate resources to adequately maintain infrastructure and equipment, however, both traditional and conventional practices indicate that decisions to refurbish, replace, renovate, or upgrade infrastructure and/or equipment tend to be based on...

Prioritisation of responsive maintenance tasks via machine learning based inference

Eirini Konstantinou, Ajith Parlikad, Alex Wong & Charlotte Broom
Maintenance task prioritization is essential for allocating resources. It is estimated that almost 1/3 of the maintenance cost is wasted to unnecessary activities. Task prioritization is based on risk assessment that takes into account the probability of failure and the criticality of an asset. The criticality analysis is defined by the asset owner based on several parameters, among them safety, downtime cost, productivity, whilst the probability of failure is determined based on deterioration models, regular...

Recurrent Neural Networks for real-time distributed collaborative prognostics

Ajith Parlikad, Adrià Salvador Palau, Kshitij Bakliwal, Maharshi Harshadbhai Dhada & Tim Pearce
We present the first steps towards real-time distributed collaborative prognostics enabled by an implementation of the Weibull Time To Event - Recurrent Neural Network (WTTE-RNN) algorithm. In our system, assets determine their time to failure (TTF) in real-time according to an asset-specific model that is obtained in collaboration with other similar assets in the asset fleet. The presented approach builds on the emergent field of similarity analysis in asset management, and extends it to distributed...

Exploiting traffic data to improve asset management and citizen quality of life

Adrià Salvador Palau, Jon Roozenbeek & Ajith Parlikad
The main goal of this project was to demonstrate how large data sources such as Google Maps can be used to inform transportation-related asset management decisions. Specifically, we investigated how the interdependence between infrastructures and assets can be studied using transportation data and heat maps. This involves linking the effect of disruptions in lower-order assets to travel accessibility to private and public infrastructure. In order to demonstrate the viability of our approach, we conducted 5...

Basal mitophagy is widespread in Drosophila but minimally affected by loss of Pink1 or parkin.

Juliette J Lee, Alvaro Sanchez-Martinez, Aitor Martinez Zarate, Cristiane Benincá, Ugo Mayor, Michael J Clague & Alex Whitworth
The Parkinson's disease factors PINK1 and parkin are strongly implicated in stress-induced mitophagy in vitro, but little is known about their impact on basal mitophagy in vivo. We generated transgenic Drosophila melanogaster expressing fluorescent mitophagy reporters to evaluate the impact of Pink1/parkin mutations on basal mitophagy under physiological conditions. We find that mitophagy is readily detectable and abundant in many tissues, including Parkinson's disease-relevant dopaminergic neurons. However, we did not detect mitolysosomes in flight muscle....


Felix Stahlberg, Danielle Saunders, Adria De Gispert & William Byrne
Two techniques provide the fabric of the Cambridge University Engineering Department's (CUED) entry to the WMT19 evaluation campaign: elastic weight consolidation (EWC) and different forms of language modelling (LMs). We report substantial gains by fine-tuning very strong baselines on former WMT test sets using a combination of checkpoint averaging and EWC. A sentence-level Transformer LM and a document-level LM based on a modified Transformer architecture yield further gains. As in previous years, we also extract...

A spatial mixture approach to inferring sub-ROI spatio-temporal patterns from rapid event-related fMRI data.

Yuan Shen, Stephen Mayhew, Zoe Kourtzi & Peter Tino
Previous works investigated a range of spatio-temporal models for fMRI data analysis to provide robust determination of functional region-of-interest (ROI). We present a novel spatio-temporal fMRI model that is suitable for identifying a number of distinct temporal patterns and their spatial support in the voxel space. Accordingly, fMRI signals on a single voxel are modeled as a probabilistic superposition of those temporal patterns. The spatially varying influence of individual patterns is defined in terms of...

Registration Year

  • 2016
  • 2017
  • 2018
  • 2019

Resource Types

  • Collection

Data Centers

  • University of Cambridge