Overfitting Bayesian mixture models with an unknown number of components

This paper proposes solutions to three issues pertaining to the estimation of finite mixture models with an unknown number of components: the non-identifiability induced by overfitting the number of components, the mixing limitations of standard Markov Chain Monte Carlo (MCMC) sampling techniques, and the related label switching problem. An overfitting approach is used to estimate the number of components in a finite mixture model via a Zmix algorithm. Zmix provides a bridge between multidimensional samplers and test based estimation methods, whereby priors are chosen to encourage extra groups to have weights approaching zero. MCMC sampling is made possible by the implementation of prior parallel tempering, an extension of parallel tempering. Zmix can accurately estimate the number of components, posterior parameter estimates and allocation probabilities given a sufficiently large sample size. The results will reflect uncertainty in the final model and will report the range of possible candidate models and their respective estimated probabilities from a single run. Label switching is resolved with a computationally light-weight method, Zswitch, developed for overfitted mixtures by exploiting the intuitiveness of allocation-based relabelling algorithms and the precision of label-invariant loss functions. Four simulation studies are included to illustrate Zmix and Zswitch, as well as three case studies from the literature. All methods are available as part of the R package Zmix, which can currently be applied to univariate Gaussian mixture models.

When film, fashion & interiors collide: Designing the After Darkly Graduate Fashion Show

In 2013 QUT Interior Design and Fashion Disciplines partnered to design the Catwalk for the QUT After Darkly Graduate Fashion Show. The ephemeral work (catwalk canopy and cinematic affects) was developed through collaboration between the authors based upon an undergraduate interior design unit ‘Filmic Interiors’ in which students were tasked with designing a fashion show. Filmic Interiors exploited the potential of film to influence, understand, and develop novel interior spaces through consideration of mise-en-scene, cinematic effects and atmospheric design strategies engaged by key film directors Jean Pierre Jeunet and Darren Aronofsky. The design outcome represents a hybridisation of student design proposals, contemplating both film and emerging collections from graduate fashion students. The work explored a number of iterations each testing material qualities and immaterial cinematic affects, as a means to develop new space. The process was led by experimentation undertaken by the designers through previous studio explorations surrounding the theme of ‘Strange Space’ and design practice ‘Making Strange’(Lindquist & Pytel, 2012). In doing so, the work paralleled the material formations of ‘obsessive collections’ and ‘making do’ evident in Jeunet’s scenography, rendering uncanny hybrid space (Ezra, 2008). Evocation of the immaterial found in much of director Aronofsky’s work, also became critical in the atmospheric experience intended for the show. This paper explores the process of collaboration and material experimentation in design, approached through a filmic lens. It provides insight into what happens when one enters into what can be termed an ‘ecology of production’, whereby the experimental making becomes the collaborative agent between designers, disciplines, and between stage and spectators. Finally it underlines the importance of ‘finding the work’ through material making and testing rather than through more controlled formalistic responses.

Domain Adaptation on the Statistical Manifold

In this paper, we tackle the problem of unsupervised domain adaptation for classification. In the unsupervised scenario where no labeled samples from the target domain are provided, a popular approach consists in transforming the data such that the source and target distributions be- come similar. To compare the two distributions, existing approaches make use of the Maximum Mean Discrepancy (MMD). However, this does not exploit the fact that prob- ability distributions lie on a Riemannian manifold. Here, we propose to make better use of the structure of this man- ifold and rely on the distance on the manifold to compare the source and target distributions. In this framework, we introduce a sample selection method and a subspace-based method for unsupervised domain adaptation, and show that both these manifold-based techniques outperform the cor- responding approaches based on the MMD. Furthermore, we show that our subspace-based approach yields state-of- the-art results on a standard object recognition benchmark.

Where the action isn't, that's where it is

Contribution to ARI Remix. ARI remix is a three-year digital humanities, artist interviews and oral history project collecting and presenting memories of Australian Artist-run culture in Queensland, New South Wales and the Australian Capital Territory between 1980 and 2000. Its focus is fleshing out and illuminating the ephemeral and neglected histories of the many lively and socially engaged artistic scenes along the east coast of Australia during the last two decades of the 20th century.

Recent developments in DNA evidence

DNA evidence has made a significant contribution to criminal investigations in Australia and around the world since it was widely adopted in the 1990s (Gans & Urbas 2002). The direct matching of DNA profiles, such as comparing one obtained from a crime scene with one obtained from a suspect or database, remains a widely used technique in criminal investigations. A range of new DNA profiling techniques continues to be developed and applied in criminal investigations around the world (Smith & Urbas 2012). This paper is the third in a series by the Australian Institute of Criminology (AIC) on DNA evidence. The first, published in 1990 when the technology was in its relative infancy, outlined the scientific background for DNA evidence, considered early issues such as scientific reliability and privacy and described its application in early criminal cases (Easteal & Easteal 1990). The second, published in 2002, expanded on the scientific background and discussed a significant number of Australian cases in a 12-year period, illustrating issues that had arisen in investigations, at trial and in the use of DNA in the review of convictions and acquittals (Gans & Urbas 2002). There have been some significant developments in the science and technology behind DNA evidence in the 13 years since 2002 that have important implications for law enforcement and the legal system. These are discussed through a review of relevant legal cases and the latest empirical evidence. This paper is structured in three sections. The first examines the scientific techniques and how they have been applied in police investigations, drawing on a number of recent cases to illustrate them. The second considers empirical research evaluating DNA evidence and databases and the impact DNA has on investigative and court outcomes. The final section discusses significant cases that establish legal precedent relating to DNA evidence in criminal trials where significant issues have arisen or new techniques have been applied that have not yet been widely discussed in the literature. The paper concludes by reflecting on implications for policy and practice.

Multiple instance dictionary learning for activity representation

This paper presents an effective feature representation method in the context of activity recognition. Efficient and effective feature representation plays a crucial role not only in activity recognition, but also in a wide range of applications such as motion analysis, tracking, 3D scene understanding etc. In the context of activity recognition, local features are increasingly popular for representing videos because of their simplicity and efficiency. While they achieve state-of-the-art performance with low computational requirements, their performance is still limited for real world applications due to a lack of contextual information and models not being tailored to specific activities. We propose a new activity representation framework to address the shortcomings of the popular, but simple bag-of-words approach. In our framework, first multiple instance SVM (mi-SVM) is used to identify positive features for each action category and the k-means algorithm is used to generate a codebook. Then locality-constrained linear coding is used to encode the features into the generated codebook, followed by spatio-temporal pyramid pooling to convey the spatio-temporal statistics. Finally, an SVM is used to classify the videos. Experiments carried out on two popular datasets with varying complexity demonstrate significant performance improvement over the base-line bag-of-feature method.

Appalling behaviour

'Appalling Behaviour' is a critically acclaimed contemporary Australian monologue, written by AWGIE Award winning playwright, Stephen House. This production, directed and creatively adapted by Shane Pike, was presented at the Brisbane Powerhouse in February 2016, as part of Queensland's LGBTIQ festival, Melt. This adaptation of the work experimented with notions of gender, taking the original script and manipulating character and scene to investigate expressions of identity beyond the traditional notions of binary gender-norms. To this end, the sole character (and actor) was (re)presented as a homeless bi-sexual queen with the aim of inferring that gender un/ab-normative characters can exist not only as disruptors/comments on/agitators of traditional expectations of performed gender (both onstage and off), but can also exist as accepted characters in and of themselves. Put simply: can a bi-sexual queen just be an actor/character in a play, or do all gender extra-normative characters inherently exist as political, social and cultural challengers? If so, why is this the case and should we be aiming for this kind of character to be an accepted part of the performative fabric, seamless and fitting within any onstage situation and play (why can't Willy Loman, King Lear or Nora be gender non-normative), or should such (re)presentations always exist as 'different'? Is it time for individual expressions of gender to just 'be' and be accepted as givens, or are we not quite there yet?

NKT cells from normal and tumor-bearing human livers are phenotypically and functionally distinct from murine NKT cells

A major group of murine NK T (NKT) cells express an invariant Vα14Jα18 TCR α-chain specific for glycolipid Ags presented by CD1d. Murine Vα14Jα18+ account for 30–50% of hepatic T cells and have potent antitumor activities. We have enumerated and characterized their human counterparts, Vα24Vβ11+ NKT cells, freshly isolated from histologically normal and tumor-bearing livers. In contrast to mice, human NKT cells are found in small numbers in healthy liver (0.5% of CD3+ cells) and blood (0.02%). In contrast to those in blood, most hepatic Vα24+ NKT cells express the Vβ11 chain. They include CD4+, CD8+, and CD4−CD8− cells, and many express the NK cell markers CD56, CD161, and/or CD69. Importantly, human hepatic Vα24+ T cells are potent producers of IFN-γ and TNF-α, but not IL-2 or IL-4, when stimulated pharmacologically or with the NKT cell ligand, α-galactosylceramide. Vα24+Vβ11+ cell numbers are reduced in tumor-bearing compared with healthy liver (0.1 vs 0.5%; p < 0.04). However, hepatic cells from cancer patients and healthy donors release similar amounts of IFN-γ in response to α-galactosylceramide. These data indicate that hepatic NKT cell repertoires are phenotypically and functionally distinct in humans and mice. Depletions of hepatic NKT cell subpopulations may underlie the susceptibility to metastatic liver disease.

CD1 expression and CD1-restricted T cell activity in normal and tumour-bearing human liver

CD1d-restricted natural killer T (NKT) cells expressing invariant Valpha14Jalpha18 T cell receptor alpha-chains are abundant in murine liver and are implicated in the control of malignancy, infection and autoimmunity. Invariant NKT cells have potent anti-metastatic effects in mice and phase I clinical trials involving their homologues in humans are ongoing. However, invariant NKT cells are less abundant in human liver ( approximately 0.5% of hepatic T cells) than in murine liver (up to 50%) and it is not known if other hepatic T cells are CD1-restricted. We have examined expression of CD1a, CD1b, CD1c and CD1d mRNA and protein in human liver and evaluated the reactivity of mononuclear cells (MNC) from histologically normal and tumour-bearing human liver specimens against these CD1 isoforms. Messenger RNA for all CD1 isotypes was detectable in all liver samples. CD1c and CD1d were expressed at the protein level by hepatic MNC. CD1d, only, was detectable at the cell surface, but CD1c and CD1d were found at an intracellular location in significant numbers of liver MNC. CD1b was not expressed by MNC from healthy livers but was detectable within MNC in all tumour samples tested. Hepatic T cells exhibited reactivity against C1R cells expressing transfected CD1c and CD1d, but neither CD1a nor CD1b. These cells secreted interferon-gamma (IFN-gamma) but not interleukin-4 (IL-4) upon stimulation. In contrast, similar numbers of peripheral T cells released 13- and 16-fold less IFN-gamma in response to CD1c and CD1d, respectively. CD1c and CD1d expression and T cell reactivity were not altered in tumour-bearing liver specimens compared to histologically normal livers. These data suggest that, in addition to invariant CD1d-restricted NKT cells, autoreactive T cells that recognise CD1c and CD1d and release inflammatory cytokines are abundant in human liver.

Fast forward to the end: Stories of the cassette magazine

Although they made up a significant share of the retail music market from the 60s to the 90s, and were often used as promotional and marketing tools, cassettes became re-configured in the 80s as an integral component of cassette magazines. Binding sound, music, talk, text, visual art and design, they were a truly innovative interdisciplinary form. This paper explores the history of these artefacts with particular emphasis on the Brisbane underground music scene of the late 70s and 80s, and discusses their significance in as a bridge between the music scene and art scenes of this period.

It's not the heat, it's the humidity

The performance aspect of a PhD thesis. In 1976, The Saints released “I’m Stranded” and Brisbane was suddenly on the map. People starting talking about Brisbane bands in the same way they had talked about Liverpool bands or would talk about Seattle bands. The history of the Brisbane music scene has been constructed and interpreted by both journalists (at the time and since) and academics (in retrospect). These histories and analyses have come some way to painting a picture of a time and place that is somewhat mythic. As a significant member of this scene, my memory stores a different picture, different shadings: a vibrant social scene with cultural by-products (music, art, film, fashion). By gathering the ephemera of the time and embellishing with a series of interviews with some Brisbane musicians, I will be growing new, untold stories, grounded in shared experience and understanding. The interview footage, and the bounty of ephemera that continues to be unearthed, will be thrown together in the style of the times and presented in public as a live documentary filled with the faces, voices and music on which these times were built. The performances are the Creative Practice component of the PhD, and are the result of a curatorial process which will be examined in the exegetic component of the thesis due for delivery in mid 2016.

Ma Ma Ma Mad

Ma Ma Ma Mad is an autobiographical work, written and performed by Singaporean-Australian theatre maker Merlynn Tong. This production, presented at the Brisbane Powerhouse in December 2015, was a multi-genre work incorporating aspects of Butoh, physical theatre, cabaret and contemporary monologue. More than an experiment in mixed performative forms, however, this particular production was also an exercise in inter-cultural collaboration as well as gender in (and of) performance. Heavily influenced by the creator's experiences growing up in urban Southeast Asia, the director's specialisation in contemporary Australian theatre and experience telling uniquely Australian stories worked to manipulate the form in an endeavour to succinctly speak to local audiences, without pandering to entrenched stereotypes or diluting the underlying Chinese-Singaporean themes. The success of this production was also somewhat of a personal challenge for the creatives, after being told by some of Brisbane's most influential theatre venues and festivals that they would rather not support the work because a) it was a one woman show, and b) it was a one woman show about an Asian woman; and therefore would not sell well. One very influential local producer even said that he already had a one-woman show about an Asian person programmed, so he couldn't possibly program another. Operating in such a biased and out-of-touch artistic environment was seen as an easy challenge for the artists involved, which resulted in a highly successful and critically acclaimed sell-out run of Ma Ma Ma Mad, followed by offers to tour the work nationally and internationally. As such, this production also stands as a practical example of the ingrained and patriarchal structures of the Australian arts scene, and how art can work to break down the very barriers that it has helped to construct through a lack of vision and diversity amongst its leaders.

Using multi-label classification for acoustic pattern detection and assisting bird species surveys

Acoustics is a rich source of environmental information that can reflect the ecological dynamics. To deal with the escalating acoustic data, a variety of automated classification techniques have been used for acoustic patterns or scene recognition, including urban soundscapes such as streets and restaurants; and natural soundscapes such as raining and thundering. It is common to classify acoustic patterns under the assumption that a single type of soundscapes present in an audio clip. This assumption is reasonable for some carefully selected audios. However, only few experiments have been focused on classifying simultaneous acoustic patterns in long-duration recordings. This paper proposes a binary relevance based multi-label classification approach to recognise simultaneous acoustic patterns in one-minute audio clips. By utilising acoustic indices as global features and multilayer perceptron as a base classifier, we achieve good classification performance on in-the-field data. Compared with single-label classification, multi-label classification approach provides more detailed information about the distributions of various acoustic patterns in long-duration recordings. These results will merit further biodiversity investigations, such as bird species surveys.

Stationary processes for object detection and non-rigid structure-from-motion

Stationary processes are random variables whose value is a signal and whose distribution is invariant to translation in the domain of the signal. They are intimately connected to convolution, and therefore to the Fourier transform, since the covariance matrix of a stationary process is a Toeplitz matrix, and Toeplitz matrices are the expression of convolution as a linear operator. This thesis utilises this connection in the study of i) efficient training algorithms for object detection and ii) trajectory-based non-rigid structure-from-motion.