Human action recognition from video sequences

This PhD research has proposed new machine learning techniques to improve human action recognition based on local features. Several novel video representation and classification techniques have been proposed to increase the performance with lower computational complexity. The major contributions are the construction of new feature representation techniques, based on advanced machine learning techniques such as multiple instance dictionary learning, Latent Dirichlet Allocation (LDA) and Sparse coding. A Binary-tree based classification technique was also proposed to deal with large amounts of action categories. These techniques are not only improving the classification accuracy with constrained computational resources but are also robust to challenging environmental conditions. These developed techniques can be easily extended to a wide range of video applications to provide near real-time performance.

A combination of local inflammation and central memory T cells potentiates immunotherapy in the skin

Adoptive T cell therapy uses the specificity of the adaptive immune system to target cancer and virally infected cells. Yet the mechanism and means by which to enhance T cell function are incompletely described, especially in the skin. In this study, we use a murine model of immunotherapy to optimize cell-mediated immunity in the skin. We show that in vitro - derived central but not effector memory-like T cells bring about rapid regression of skin-expressing cognate Ag as a transgene in keratinocytes. Local inflammation induced by the TLR7 receptor agonist imiquimod subtly yet reproducibly decreases time to skin graft rejection elicited by central but not effector memory T cells in an immunodeficient mouse model. Local CCL4, a chemokine liberated by TLR7 agonism, similarly enhances central memory T cell function. In this model, IL-2 facilitates the development in vivo of effector function from central memory but not effector memory T cells. In a model of T cell tolerogenesis, we further show that adoptively transferred central but not effector memory T cells can give rise to successful cutaneous immunity, which is dependent on a local inflammatory cue in the target tissue at the time of adoptive T cell transfer. Thus, adoptive T cell therapy efficacy can be enhanced if CD8+ T cells with a central memory T cell phenotype are transferred, and IL-2 is present with contemporaneous local inflammation. Copyright © 2012 by The American Association of Immunologists, Inc.

Steady-state dendritic cells expressing cognate antigen terminate memory CD8+ T-cell responses

Antigen stimulation of naive T cells in conjunction with strong costimulatory signals elicits the generation of effector and memory populations. Such terminal differentiation transforms naive T cells capable of differentiating along several terminal pathways in response to pertinent environmental cues into cells that have lost developmental plasticity and exhibit heightened responsiveness. Because these cells exhibit little or no need for the strong costimulatory signals required for full activation of naive T cells, it is generally considered memory and effector T cells are released from the capacity to be inactivated. Here, we show that steadystate dendritic cells constitutively presenting an endogenously expressed antigen inactivate fully differentiated memory and effector CD8+ T cells in vivo through deletion and inactivation. These findings indicate that fully differentiated effector and memory T cells exhibit a previously unappreciated level of plasticity and provide insight into how memory and effector T-cell populations may be regulated. © 2008 by The American Society of Hematology.

Targeting antigen to diverse APCs inactivates memory CD8+ T cells without eliciting tissue-destructive effector function

Memory T cells develop early during the preclinical stages of autoimmune diseases and have traditionally been considered resistant to tolerance induction. As such, they may represent a potent barrier to the successful immunotherapy of established autoimmune diseases. It was recently shown that memory CD8+ T cell responses are terminated when Ag is genetically targeted to steady-state dendritic cells. However, under these conditions, inactivation of memory CD8+ T cells is slow, allowing transiently expanded memory CD8+ T cells to exert tissue-destructive effector function. In this study, we compared different Ag-targeting strategies and show, using an MHC class II promoter to drive Ag expression in a diverse range of APCs, that CD8+ memory T cells can be rapidly inactivated by MHC class II+ hematopoietic APCs through a mechanism that involves a rapid and sustained downregulation of TCR, in which the effector response of CD8+ memory cells is rapidly truncated and Ag-expressing target tissue destruction is prevented. Our data provide the first demonstration that genetically targeting Ag to a broad range of MHC class II+ APC types is a highly efficient way to terminate memory CD8+ T cell responses to prevent tissue-destructive effector function and potentially established autoimmune diseases. Copyright © 2010 by The American Association of Immunologists, Inc.

Elliptical summary randomisation for sensor-based human activity recognition

Many conventional statistical machine learning al- gorithms generalise poorly if distribution bias ex- ists in the datasets. For example, distribution bias arises in the context of domain generalisation, where knowledge acquired from multiple source domains need to be used in a previously unseen target domains. We propose Elliptical Summary Randomisation (ESRand), an efficient domain generalisation approach that comprises of a randomised kernel and elliptical data summarisation. ESRand learns a domain interdependent projection to a la- tent subspace that minimises the existing biases to the data while maintaining the functional relationship between domains. In the latent subspace, ellipsoidal summaries replace the samples to enhance the generalisation by further removing bias and noise in the data. Moreover, the summarisation enables large-scale data processing by significantly reducing the size of the data. Through comprehensive analysis, we show that our subspace-based approach outperforms state-of-the-art results on several activity recognition benchmark datasets, while keeping the computational complexity significantly low.

Memory and Gender in Medieval Europe 900-1200

Review of Memory and Gender in Medieval Europe, 900-1200 by Elizabeth van Houts (Toronto UP, 1999).