Gaussian process approximations of stochastic differential equations

Stochastic differential equations arise naturally in a range of contexts, from financial to environmental modeling. Current solution methods are limited in their representation of the posterior process in the presence of data. In this work, we present a novel Gaussian process approximation to the posterior measure over paths for a general class of stochastic differential equations in the presence of observations. The method is applied to two simple problems: the Ornstein-Uhlenbeck process, of which the exact solution is known and can be compared to, and the double-well system, for which standard approaches such as the ensemble Kalman smoother fail to provide a satisfactory result. Experiments show that our variational approximation is viable and that the results are very promising as the variational approximate solution outperforms standard Gaussian process regression for non-Gaussian Markov processes.

Optical biochemical sensors based on long-period fibre gratings UV-inscribed in D-fibre with enhanced sensitivity by HF etching process

We report a new concept of biochemical sensor device based on long-period grating structures UV-inscribed in D-fiber. The surrounding-medium refractive index sensitivity of the devices has been enhanced significantly by a hydrofluoric acid etching process. The devices have been used to measure the sugar concentrations showing clearly an encoding relation between the chemical concentration and the grating spectral response, demonstrating their capability for potential biochemical sensing applications.

Monitoring of deployment process of shape memory polymers for morphing structures with embedded fibre Bragg grating sensors

This article demonstrates the use of embedded fibre Bragg gratings as vector bending sensor to monitor two-dimensional shape deformation of a shape memory polymer plate. The shape memory polymer plate was made by using thermal-responsive epoxy-based shape memory polymer materials, and the two fibre Bragg grating sensors were orthogonally embedded, one on the top and the other on the bottom layer of the plate, in order to measure the strain distribution in both longitudinal and transverse directions separately and also with temperature reference. When the shape memory polymer plate was bent at different angles, the Bragg wavelengths of the embedded fibre Bragg gratings showed a red-shift of 50 pm/°caused by the bent-induced tensile strain on the plate surface. The finite element method was used to analyse the stress distribution for the whole shape recovery process. The strain transfer rate between the shape memory polymer and optical fibre was also calculated from the finite element method and determined by experimental results, which was around 0.25. During the experiment, the embedded fibre Bragg gratings showed very high temperature sensitivity due to the high thermal expansion coefficient of the shape memory polymer, which was around 108.24 pm/°C below the glass transition temperature (Tg) and 47.29 pm/°C above Tg. Therefore, the orthogonal arrangement of the two fibre Bragg grating sensors could provide a temperature compensation function, as one of the fibre Bragg gratings only measures the temperature while the other is subjected to the directional deformation. © The Author(s) 2013.

Detecting Precipitation Climate Changes: An Approach Based on a Stochastic Daily Precipitation Model

2002 Mathematics Subject Classification: 62M10.

Stories from the Inside:The Prisons Memory Archive

A close textual reading of the online-available recording of ex-prison officer, John Heatherington, who is filmed as he returns to the site of the Maze and Long Kesh Prison, which was the largest prison oeperating during the Troubles in Northern Ireland. Using the empty site as stimulant for his memory, John recalls the tension, violence and dark humour of the period, while also showing exceptional empathy for all those, including prisoners, who passed through the prison system.The interview is taken from the Prisons Memory Archive (www.prisonsmemoryarchive.com) and, as director of the PMA, I have a particular insight to the recording process and John's special contribution to it.

The stochastic geometric machine model

This paper introduces the stochastic version of the Geometric Machine Model for the modelling of sequential, alternative, parallel (synchronous) and nondeterministic computations with stochastic numbers stored in a (possibly infinite) shared memory. The programming language L(D! 1), induced by the Coherence Space of Processes D! 1, can be applied to sequential and parallel products in order to provide recursive definitions for such processes, together with a domain-theoretic semantics of the Stochastic Arithmetic. We analyze both the spacial (ordinal) recursion, related to spacial modelling of the stochastic memory, and the temporal (structural) recursion, given by the inclusion relation modelling partial objects in the ordered structure of process construction.

TIMESCAPES: an architectural process of memory and adaptation

What if the architectural process of making could incorporate time? All designers who impact the physical environment- consciously and unconsciously are gatekeepers of the past, commentators of the present, and speculators of the future. This project proposes the creation of architecture and adaptive public space that looks to historical memories, foster present day cultural formation, and new alternative visions for the city of the future. The thesis asks what it means to design for stasis and change in a variety of scales- urban, architectural, and detail and arrives at a speculated new neighborhood, institutional buildings, and landscape. Central to this project is the idea of the architect as archeologist, anthropologist, and artist. The project focuses on a rapidly changing part of the city of Fort Worth, Texas and assigns a multipurpose institutional buildings and public space as a method of investigation. The thesis hopes to further architectural discourse about into the role of architecture in the preservation of memory, adaptive potential of public spaces, and the role of time in architecture.

Datasets, features, learning, and models in visual recognition

Visual recognition is a fundamental research topic in computer vision. This dissertation explores datasets, features, learning, and models used for visual recognition. In order to train visual models and evaluate different recognition algorithms, this dissertation develops an approach to collect object image datasets on web pages using an analysis of text around the image and of image appearance. This method exploits established online knowledge resources (Wikipedia pages for text; Flickr and Caltech data sets for images). The resources provide rich text and object appearance information. This dissertation describes results on two datasets. The first is Berg’s collection of 10 animal categories; on this dataset, we significantly outperform previous approaches. On an additional set of 5 categories, experimental results show the effectiveness of the method. Images are represented as features for visual recognition. This dissertation introduces a text-based image feature and demonstrates that it consistently improves performance on hard object classification problems. The feature is built using an auxiliary dataset of images annotated with tags, downloaded from the Internet. Image tags are noisy. The method obtains the text features of an unannotated image from the tags of its k-nearest neighbors in this auxiliary collection. A visual classifier presented with an object viewed under novel circumstances (say, a new viewing direction) must rely on its visual examples. This text feature may not change, because the auxiliary dataset likely contains a similar picture. While the tags associated with images are noisy, they are more stable when appearance changes. The performance of this feature is tested using PASCAL VOC 2006 and 2007 datasets. This feature performs well; it consistently improves the performance of visual object classifiers, and is particularly effective when the training dataset is small. With more and more collected training data, computational cost becomes a bottleneck, especially when training sophisticated classifiers such as kernelized SVM. This dissertation proposes a fast training algorithm called Stochastic Intersection Kernel Machine (SIKMA). This proposed training method will be useful for many vision problems, as it can produce a kernel classifier that is more accurate than a linear classifier, and can be trained on tens of thousands of examples in two minutes. It processes training examples one by one in a sequence, so memory cost is no longer the bottleneck to process large scale datasets. This dissertation applies this approach to train classifiers of Flickr groups with many group training examples. The resulting Flickr group prediction scores can be used to measure image similarity between two images. Experimental results on the Corel dataset and a PASCAL VOC dataset show the learned Flickr features perform better on image matching, retrieval, and classification than conventional visual features. Visual models are usually trained to best separate positive and negative training examples. However, when recognizing a large number of object categories, there may not be enough training examples for most objects, due to the intrinsic long-tailed distribution of objects in the real world. This dissertation proposes an approach to use comparative object similarity. The key insight is that, given a set of object categories which are similar and a set of categories which are dissimilar, a good object model should respond more strongly to examples from similar categories than to examples from dissimilar categories. This dissertation develops a regularized kernel machine algorithm to use this category dependent similarity regularization. Experiments on hundreds of categories show that our method can make significant improvement for categories with few or even no positive examples.

Price Trackers Inspired by Immune Memory

In this paper we outline initial concepts for an immune inspired algorithm to evaluate price time series data. The proposed solution evolves a short term pool of trackers dynamically through a process of proliferation and mutation, with each member attempting to map to trends in price movements. Successful trackers feed into a long term memory pool that can generalise across repeating trend patterns. Tests are performed to examine the algorithm’s ability to successfully identify trends in a small data set. The influence of the long term memory pool is then examined. We find the algorithm is able to identify price trends presented successfully and efficiently.

Time-Dependent Effects of Acute Exercise-Induced Arousal on Long-Term Memory for Emotional and Neutral Stimuli

Psychological research has strongly documented the memory-enhancing effects of emotional arousal, while the effects of acute aerobic exercise on memory are not well understood. Manipulation of arousal has been shown to enhance long-term memory for emotional stimuli in a time-dependent fashion. This presents an opportunity to investigate the role of acute exercise in memory modulation. The purpose of this study was to determine the time-dependent relationship between acute exercise-induced arousal and long-term emotional memory. Participants viewed pleasant, neutral, and unpleasant images before or after completing a high-intensity session of cycling exercise. Salivary alpha-amylase, a biomarker of central norepinephrine, was measured as an indicator of arousal. No effects of exercise on recognition memory were revealed, however; a single session of high-intensity cycling increased salivary alpha-amylase. Our results also indicate that the influence of exercise on emotional responsiveness should be considered in further exploration of the memory-enhancing potential of acute exercise.

Investigating assistive technology to support memory for people with cognitive impairments

Technologies such as automobiles or mobile phones allow us to perform beyond our physical capabilities and travel faster or communicate over long distances. Technologies such as computers and calculators can also help us perform beyond our mental capabilities by storing and manipulating information that we would be unable to process or remember. In recent years there has been a growing interest in assistive technology for cognition (ATC) which can help people compensate for cognitive impairments. The aim of this thesis was to investigate ATC for memory to help people with memory difficulties which impacts independent functioning during everyday life. Chapter one argues that using both neuropsychological and human computing interaction theory and approaches is crucial when developing and researching ATC. Chapter two describes a systematic review and meta-analysis of studies which tested technology to aid memory for groups with ABI, stroke or degenerative disease. Good evidence was found supporting the efficacy of prompting devices which remind the user about a future intention at a set time. Chapter three looks at the prevalence of technologies and memory aids in current use by people with ABI and dementia and the factors that predicted this use. Pre-morbid use of technology, current use of non-tech aids and strategies and age (ABI group only) were the best predictors of this use. Based on the results, chapter four focuses on mobile phone based reminders for people with ABI. Focus groups were held with people with memory impairments after ABI and ABI caregivers (N=12) which discussed the barriers to uptake of mobile phone based reminding. Thematic analysis revealed six key themes that impact uptake of reminder apps; Perceived Need, Social Acceptability, Experience/Expectation, Desired Content and Functions, Cognitive Accessibility and Sensory/Motor Accessibility. The Perceived need theme described the difficulties with insight, motivation and memory which can prevent people from initially setting reminders on a smartphone. Chapter five investigates the efficacy and acceptability of unsolicited prompts (UPs) from a smartphone app (ForgetMeNot) to encourage people with ABI to set reminders. A single-case experimental design study evaluated use of the app over four weeks by three people with severe ABI living in a post-acute rehabilitation hospital. When six UPs were presented through the day from ForgetMeNot, daily reminder-setting and daily memory task completion increased compared to when using the app without the UPs. Chapter six investigates another barrier from chapter 4 – cognitive and sensory accessibility. A study is reported which shows that an app with ‘decision tree’ interface design (ApplTree) leads to more accurate reminder setting performance with no compromise of speed or independence (amount of guidance required) for people with ABI (n=14) compared to a calendar based interface. Chapter seven investigates the efficacy of a wearable reminding device (smartwatch) as a tool for delivering reminders set on a smartphone. Four community dwelling participants with memory difficulties following ABI were included in an ABA single case experimental design study. Three of the participants successfully used the smartwatch throughout the intervention weeks and these participants gave positive usability ratings. Two participants showed improved memory performance when using the smartwatch and all participants had marked decline in memory performance when the technology was removed. Chapter eight is a discussion which highlights the implications of these results for clinicians, researchers and designers.

Review of Two Issues at the Long-Term Commitment Facilities, Department of Juvenile Justice: 1)Safety Threat to Employees & Juveniles; and 2)Event Reporting Process Integrity Allegations

The Legislative Oversight Committee of the South Carolina House of Representatives, referred allegations pertaining to the Department of Juvenile Justice (DJJ) which were generated during its ongoing oversight study of DJJ. Specifically, the safety issues focused on lack of control; lack of trust; and lack of adequate staffing. This review’s scope and objectives were: Investigate specific complainant allegations of DJJ employees underreporting, misreporting, or destroying ERs; Review the efficiency and effectiveness of DJJ’s event reporting process and follow-up on anomalies or potential patterns of systemic underreporting, misreporting, or missing ERs; and Assess juvenile and employee safety conditions through interviewing a cross-section of relevant employees, record review, and possibly an employee survey.

How long is the memory of the US stock market?

The Efficient Market Hypothesis (EMH), one of the most important hypothesis in financial economics, argues that return rates have no memory (correlation) which implies that agents cannot make abnormal profits in financial markets, due to the possibility of arbitrage operations. With return rates for the US stock market, we corroborate the fact that with a linear approach, return rates do not show evidence of correlation. However, linear approaches might not be complete or global, since return rates could suffer from nonlinearities. Using detrended cross-correlation analysis and its correlation coefficient, a methodology which analyzes long-range behavior between series, we show that the long-range correlation of return rates only ends in the 149th lag, which corresponds to about seven months. Does this result undermine the EMH?