Incremental learning from stream data

Recent years have witnessed an incredibly increasing interest in the topic of incremental learning. Unlike conventional machine learning situations, data flow targeted by incremental learning becomes available continuously over time. Accordingly, it is desirable to be able to abandon the traditional assumption of the availability of representative training data during the training period to develop decision boundaries. Under scenarios of continuous data flow, the challenge is how to transform the vast amount of stream raw data into information and knowledge representation, and accumulate experience over time to support future decision-making process. In this paper, we propose a general adaptive incremental learning framework named ADAIN that is capable of learning from continuous raw data, accumulating experience over time, and using such knowledge to improve future learning and prediction performance. Detailed system level architecture and design strategies are presented in this paper. Simulation results over several real-world data sets are used to validate the effectiveness of this method.

Investigating the efficiency of fitting Coxian phase-type distributions to health care data

Coxian phase-type distributions are becoming a popular means of representing survival times within a health care environment. They are favoured as they show a distribution as a system of phases and can allow for an easy visual representation of the rate of flow of patients through a system. Difficulties arise, however, in determining the parameter estimates of the Coxian phase-type distribution. This paper examines ways of making the fitting of the Coxian phase-type distribution less cumbersome by outlining different software packages and algorithms available to perform the fit and assessing their capabilities through a number of performance measures. The performance measures rate each of the methods and help in identifying the more efficient. Conclusions drawn from these performance measures suggest SAS to be the most robust package. It has a high rate of convergence in each of the four example model fits considered, short computational times, detailed output, convergence criteria options, along with a succinct ability to switch between different algorithms.

Max separation clustering for feature extraction from Optical Emission Spectroscopy Data

This paper proposes max separation clustering (MSC), a new non-hierarchical clustering method used for feature extraction from optical emission spectroscopy (OES) data for plasma etch process control applications. OES data is high dimensional and inherently highly redundant with the result that it is difficult if not impossible to recognize useful features and key variables by direct visualization. MSC is developed for clustering variables with distinctive patterns and providing effective pattern representation by a small number of representative variables. The relationship between signal-to-noise ratio (SNR) and clustering performance is highlighted, leading to a requirement that low SNR signals be removed before applying MSC. Experimental results on industrial OES data show that MSC with low SNR signal removal produces effective summarization of the dominant patterns in the data.

A Data Model for Content Modelling of Temporal Media

This paper describes a data model for content representation of temporal media in an IP based sensor network. The model is formed by introducing the idea of semantic-role from linguistics into the underlying concepts of formal event representation with the aim of developing a common event model. The architecture of a prototype system for a multi camera surveillance system, based on the proposed model is described. The important aspects of the proposed model are its expressiveness, its ability to model content of temporal media, and its suitability for use with a natural language interface. It also provides a platform for temporal information fusion, as well as organizing sensor annotations by help of ontologies.

Simplifying the representation of complex free-energy landscapes using sketch-map

A new scheme, sketch-map, for obtaining a low-dimensional representation of the region of phase space explored during an enhanced dynamics simulation is proposed. We show evidence, from an examination of the distribution of pairwise distances between frames, that some features of the free-energy surface are inherently high-dimensional. This makes dimensionality reduction problematic because the data does not satisfy the assumptions made in conventional manifold learning algorithms We therefore propose that when dimensionality reduction is performed on trajectory data one should think of the resultant embedding as a quickly sketched set of directions rather than a road map. In other words, the embedding tells one about the connectivity between states but does not provide the vectors that correspond to the slow degrees of freedom. This realization informs the development of sketch-map, which endeavors to reproduce the proximity information from the high-dimensionality description in a space of lower dimensionality even when a faithful embedding is not possible.

Laughter Induction Techniques Suitable for Generating Motion Capture Data of Laughter Associated Body Movements

Laughter is a frequently occurring social signal and an important part of human non-verbal communication. However it is often overlooked as a serious topic of scientific study. While the lack of research in this area is mostly due to laughter’s non-serious nature, it is also a particularly difficult social signal to produce on demand in a convincing manner; thus making it a difficult topic for study in laboratory settings. In this paper we provide some techniques and guidance for inducing both hilarious laughter and conversational laughter. These techniques were devised with the goal of capturing mo- tion information related to laughter while the person laughing was either standing or seated. Comments on the value of each of the techniques and general guidance as to the importance of atmosphere, environment and social setting are provided.

Robust Multimodal Person Identification With Limited Training Data

This paper presents a novel method of audio-visual feature-level fusion for person identification where both the speech and facial modalities may be corrupted, and there is a lack of prior knowledge about the corruption. Furthermore, we assume there are limited amount of training data for each modality (e.g., a short training speech segment and a single training facial image for each person). A new multimodal feature representation and a modified cosine similarity are introduced to combine and compare bimodal features with limited training data, as well as vastly differing data rates and feature sizes. Optimal feature selection and multicondition training are used to reduce the mismatch between training and testing, thereby making the system robust to unknown bimodal corruption. Experiments have been carried out on a bimodal dataset created from the SPIDRE speaker recognition database and AR face recognition database with variable noise corruption of speech and occlusion in the face images. The system's speaker identification performance on the SPIDRE database, and facial identification performance on the AR database, is comparable with the literature. Combining both modalities using the new method of multimodal fusion leads to significantly improved accuracy over the unimodal systems, even when both modalities have been corrupted. The new method also shows improved identification accuracy compared with the bimodal systems based on multicondition model training or missing-feature decoding alone.

Low-rank representation based action recognition

Human action recognition is an important problem in computer vision, which has been applied to many applications. However, how to learn an accurate and discriminative representation of videos based on the features extracted from videos still remains to be a challenging problem. In this paper, we propose a novel method named low-rank representation based action recognition to recognize human actions. Given a dictionary, low-rank representation aims at finding the lowestrank representation of all data, which can capture the global data structures. According to its characteristics, low-rank representation is robust against noises. Experimental results demonstrate the effectiveness of the proposed approach on several publicly available datasets.

The List_Tree - an innovative solution to customer portfolio representation

The scale of BT's operations necessitates the use of very large scale computing systems, and the storage and management of large volumes of data. Customer product portfolios are an important form of data which can be difficult to store in a space efficient way. The difficulties arise from the inherently structured form of product portfolios, and the fact that they change over time as customers add or remove products. This paper introduces a new data-modelling abstraction called the List_Tree. It has been designed specifically to support the efficient storage and manipulation of customer product portfolios, but may also prove useful in other applications with similar general requirements.

Robust bimodal person identification using face and speech with limited training data and corruption of both modalities

This paper presents a novel method of audio-visual fusion for person identification where both the speech and facial modalities may be corrupted, and there is a lack of prior knowledge about the corruption. Furthermore, we assume there is a limited amount of training data for each modality (e.g., a short training speech segment and a single training facial image for each person). A new representation and a modified cosine similarity are introduced for combining and comparing bimodal features with limited training data as well as vastly differing data rates and feature sizes. Optimal feature selection and multicondition training are used to reduce the mismatch between training and testing, thereby making the system robust to unknown bimodal corruption. Experiments have been carried out on a bimodal data set created from the SPIDRE and AR databases with variable noise corruption of speech and occlusion in the face images. The new method has demonstrated improved recognition accuracy.

Similarity or Variation? Employee Representation and Consultation Approaches Amongst Liberal Market Economy Multinationals

This paper engages with the varieties of capitalism literature to investigate the employee representation and consultation approaches of liberal market economy multinational companies (MNCs), specifically Australian, British and US MNCs operating in Australia. While the literature would suggest commonality amongst these MNCs, the paper considers whether the evidence points to similarity or variation amongst liberal market headquartered MNCs. The findings contribute to filling a recognized empirical gap on MNC employment relations practice in Australia and to a better understanding of within category varieties of capitalism similarity and variation. Drawing on survey data from MNCs operating in Australia, the results demonstrated that UK-owned MNCs were the least likely to report collective structures of employee representation. Moreover, it was found that Australian MNCs were the most likely to engage in collective forms of employee representation and made less use of direct consultative mechanisms relative to their British and US counterparts. In spite of the concerted individualization of the employment relations domain over previous decades, Australian MNCs appear to have upheld more long-standing national institutional arrangements with respect to engaging with employees on a collective basis. This varies from British and US MNC approaches which denotes that our results display within category deviation in the variety of capitalism liberal market economy typology. Just as Hall and Soskice described their seminal work on liberal market economy (LME) and coordinated market economy (CME) categories as a “work-in-progress” (2001: 2), we too suggest that Australia’s evolution in the LME category, and more specifically its industrial relations system development, and the consequences for employment relations practices of its domestic MNCs, may be a work-in-progress.

Acoustic-oceanographic data fusion for prediction of oceanic acoustic elds

Tese de doutoramento, Engenharia Electrónica e Telecomunicações (Processamento de Sinal), Faculdade de Ciências e Tecnologia, Universidade do Algarve, 2014

Choosing Union Representation: The Role of Attitudes and Emotions

In the United States, most unions are recognized by a majority vote of employees through union representation elections administered by the government. Most empirical studies of individual voting behavior during union representation elections use a rational choice model. Recently, however, some have posited that voting is often influenced by emotions. We evaluate competing hypotheses about the determinants of union voting behavior by using data collected from a 2010 representation election at Delta Air Lines, a US-based company. In addition to the older rational choice framework, multiple regression results provide support for an emotional choice model. Positive feelings toward the employer are statistically significantly related to voting ‘no’ in a representation election, while positive feelings toward the union are related to a ‘yes’ vote. Effect sizes for the emotion variables were generally larger than those for the rational choice variables, suggesting that emotions may play a key role in representation election outcomes.

IMP: Imperial Metagenomics Pipeline for high-throughput sequence data

We have developed an in-house pipeline for the processing and analyses of sequence data generated during Illumina technology-based metagenomic studies of the human gut microbiota. Each component of the pipeline has been selected following comparative analysis of available tools; however, the modular nature of software facilitates replacement of any individual component with an alternative should a better tool become available in due course. The pipeline consists of quality analysis and trimming followed by taxonomic filtering of sequence data allowing reads associated with samples to be binned according to whether they represent human, prokaryotic (bacterial/archaeal), viral, parasite, fungal or plant DNA. Viral, parasite, fungal and plant DNA can be assigned to species level on a presence/absence basis, allowing – for example – identification of dietary intake of plant-based foodstuffs and their derivatives. Prokaryotic DNA is subject to taxonomic and functional analyses, with assignment to taxonomic hierarchies (kingdom, class, order, family, genus, species, strain/subspecies) and abundance determination. After de novo assembly of sequence reads, genes within samples are predicted and used to build a non-redundant catalogue of genes. From this catalogue, per-sample gene abundance can be determined after normalization of data based on gene length. Functional annotation of genes is achieved through mapping of gene clusters against KEGG proteins, and InterProScan. The pipeline is undergoing validation using the human faecal metagenomic data of Qin et al. (2014, Nature 513, 59–64). Outputs from the pipeline allow development of tools for the integration of metagenomic and metabolomic data, moving metagenomic studies beyond determination of gene richness and representation towards microbial-metabolite mapping. There is scope to improve the outputs from viral, parasite, fungal and plant DNA analyses, depending on the depth of sequencing associated with samples. The pipeline can easily be adapted for the analyses of environmental and non-human animal samples, and for use with data generated via non-Illumina sequencing platforms.

Grid structure impact in sparse point representation of derivatives

In the Sparse Point Representation (SPR) method the principle is to retain the function data indicated by significant interpolatory wavelet coefficients, which are defined as interpolation errors by means of an interpolating subdivision scheme. Typically, a SPR grid is coarse in smooth regions, and refined close to irregularities. Furthermore, the computation of partial derivatives of a function from the information of its SPR content is performed in two steps. The first one is a refinement procedure to extend the SPR by the inclusion of new interpolated point values in a security zone. Then, for points in the refined grid, such derivatives are approximated by uniform finite differences, using a step size proportional to each point local scale. If required neighboring stencils are not present in the grid, the corresponding missing point values are approximated from coarser scales using the interpolating subdivision scheme. Using the cubic interpolation subdivision scheme, we demonstrate that such adaptive finite differences can be formulated in terms of a collocation scheme based on the wavelet expansion associated to the SPR. For this purpose, we prove some results concerning the local behavior of such wavelet reconstruction operators, which stand for SPR grids having appropriate structures. This statement implies that the adaptive finite difference scheme and the one using the step size of the finest level produce the same result at SPR grid points. Consequently, in addition to the refinement strategy, our analysis indicates that some care must be taken concerning the grid structure, in order to keep the truncation error under a certain accuracy limit. Illustrating results are presented for 2D Maxwell's equation numerical solutions.