Incremental learning of temporally-coherent gaussian mixture models

In this paper we address the problem of learning Gaussian Mixture Models (GMMs) incrementally. Unlike previous approaches which universally assume that new data comes in blocks representable by GMMs which are then merged with the current model estimate, our method works for the case when novel data points arrive oneby- one, while requiring little additional memory. We keep only two GMMs in the memory and no historical data. The current fit is updated with the assumption that the number of components is fixed, which is increased (or reduced) when enough evidence for a new component is seen. This is deduced from the change from the oldest fit of the same complexity, termed the Historical GMM, the concept of which is central to our method. The performance of the proposed method is demonstrated qualitatively and quantitatively on several synthetic data sets and video sequences of faces acquired in realistic imaging conditions

Incremental learning of temporally-coherent Gaussian mixture models

In this paper we address the problem of learning Gaussian Mixture Models (GMMs) incrementally. Unlike previous approaches which universally assume that new data comes in blocks representable by GMMs which are then merged with the current model estimate, our method works for the case when novel data points arrive one- by-one, while requiring little additional memory. We keep only two GMMs in the memory and no historical data. The current fit is updated with the assumption that the number of components is fixed which is increased (or reduced) when enough evidence for a new component is seen. This is deducedfrom the change from the oldest fit of the same complexity, termed the Historical GMM, the concept of which is central to our method. The performance of the proposed method is demonstrated qualitatively and quantitatively on several synthetic data sets and video sequences of faces acquired in realistic imaging conditions.

On weak monotonicity of some mixture functions

Monotonicity with respect to all arguments is fundamental to the definition of aggregation functions. Here we study means that are not necessarily monotone. Weak monotonicity was recently proposed as a relaxation of the monotonicity condition for averaging functions. We provide results for the weak monotonicity of some importantclasses of mixture functions. With these results we are able to extend and improve the understanding of this very important class of functions.

On the weak monotonicity of Gini means and other mixture functions

Weak monotonicity was recently proposed as a relaxation of the monotonicity condition for averaging aggregation, and weakly monotone functions were shown to have desirable properties when averaging data corrupted with outliers or noise. We extended the study of weakly monotone averages by analyzing their ϕ-transforms, and we established weak monotonicity of several classes of averaging functions, in particular Gini means and mixture operators. Mixture operators with Gaussian weighting functions were shown to be weakly monotone for a broad range of their parameters. This study assists in identifying averaging functions suitable for data analysis and image processing tasks in the presence of outliers.

Cholinesterase Research Outreach Project (CROP): measuring cholinesterase activity and pesticide use in an agricultural community

BACKGROUND: Australian farmers and their workers are exposed to a wide variety of pesticides. Organophosphate (OP) insecticides are a widely used class of pesticide used for animal husbandry practices (Naphthalophos for sheep dipping, jetting and drench), crop production for pest control (Dimethoate) and in public health (Maldison for head lice). Acute poisonings with this class of insecticide are reported among agricultural workers and children around the globe, due to the inhibition of acetylcholinesterase (AChE). Less is known about chronic exposures. Regular monitoring of erythrocyte AChE will enable farmers to identify potential exposure to organophosphate insecticides and take action to reduce exposures and improve their health and safety practices. This study aims to assess and improve the integration of AChE monitoring into routine point of care health clinics, and provide farming and non-farming people with a link between their AChE activity and their household chemical and agrichemical use. METHODS/DESIGN: The research will target individuals who work on mixed farming enterprises and routinely using OPs (n = 50) and non-farmers (n = 30). Baseline data are collected regarding demographic, health conditions and behaviours, Kessler 10 (K10) scores, chemical use and personal protection. Baseline anthropometric measures include height, weight, hip and waist circumference, body fat analysis and, biochemical analysis of fasted total serum cholesterol, triglycerides, low-density cholesterol (LDL), high-density cholesterol (HDL) and blood glucose. Analysis of erythrocyte cholinesterase (EAChE) activity is also conducted using a finger prick test. Testing of EAChE is then repeated in all participants every 3 weeks for a maximum of three times over a period 10 weeks. Participants are provided with full feedback and counselling about their EAChE activity after each reading and a detailed summary provided to all participants at the completion of the study. Data will be analysed using repeated measures within a general linear model. DISCUSSION: This work will provide an evidence base and recommendations for the integration of EAChE monitoring into Australian rural health clinics, leading to research which will further quantify pesticide exposure both on the farm and in the home, highlighting the importance of sustaining and providing a safe work and home environment for farming communities. TRIAL REGISTRATION: ACTRN12613001256763.

Ultra-small algal chitosan ocular nanoparticles with iron-binding milk protein prevents the toxic effects of carbendazim pesticide

AIM: To fabricate ultra-small algal chitosan nanoparticles (US CS NPs) for efficient delivery of bovine lactoferrin (bLf) to ocular tissues through topical administration to prevent carbendazim-induced toxicity. MATERIALS & METHODS: Rat eye model was used to evaluate the in vivo biodistribution the US CS NPs and bovine eye model was used for evaluating ex vivo biodistribution. Human lens epithelial cell line (HLEB-3) model was used to evaluate the in vitro toxicity, uptake mechanism and in vitro efficacy of the synthesized bLf-US CS NPs over carbendazim-induced ocular toxicity. RESULTS: The in vivo and ex vivo biodistribution results suggest that the ultra-small CS NPs efficiently internalize into the ocular tissues within 1 h after administering topically. Ultra-small algal nanocarriers to encapsulate bioactive antioxidant bLf protein and evaluated its potential in inhibiting carbendazim-induced human lens cell apoptosis and oxidative stress. CONCLUSION: US CS NPs could be explored for their potential for delivering various ocular drugs through topical administration for other eye diseases including cataract, glaucoma and age-related macular degeneration.

Microstructure and hardness characterisation of laser coatings produced with a mixture of AISI 420 stainless steel and Fe-C-Cr-Nb-B-Mo steel alloy powders

Fe-C-Cr-Nb-B-Mo alloy powder and AISI 420 SS powder are deposited using laser cladding to increase the hardness for wear resistant applications. Mixtures from 0 to 100 wt.% were evaluated to understand the effect on the elemental composition, microstructure, phases, and microhardness. The mixture of carbon, boron and niobium in the Fe-C-Cr-Nb-B-Mo alloy powder introduces complex carbides into a Fe-based matrix of AISI 420 SS which increases its hardness. Hardness increased linearly with increasing Fe-C-Cr-Nb-B-Mo alloy, but substantial micro-cracking was observed in the clad layer at additions of 60 wt.% and above; related to a transition from a hypoeutectic alloy containing α-Fe/α' dendrites with an (Fe,Cr)2B and γ-Fe eutectic to primary and continuous carbo-borides M2B (where M represents Fe and Cr) and M23(B,C)6 carbides (where M represents Fe, Cr, Mo) with MC particles (where M represents Nb and Mo). The highest average hardness, for an alloy without micro-cracking, of 952 HV was observed in a 40 wt.% alloy. High stress abrasive scratch testing was conducted on all alloys at various loads (500, 1500, 2500 N). Alloy content was found to have a strong effect on the wear mode and the abrasive wear rate, and the presence of micro-cracks was detrimental to abrasive wear resistance.

Screening of the 'pathogen box' identifies an approved pesticide with major anthelmintic activity against the barber's pole worm

There is a substantial need to develop new medicines against parasitic diseases via public-private partnerships. Based on high throughput phenotypic screens of largely protozoal pathogens and bacteria, the Medicines for Malaria Venture (MMV) has recently assembled an open-access 'Pathogen Box' containing 400 well-curated chemical compounds. In the present study, we tested these compounds for activity against parasitic stages of the nematode Haemonchus contortus (barber's pole worm). In an optimised, whole-organism screening assay, using exsheathed third-stage (xL3) and fourth-stage (L4) larvae, we measured the inhibition of larval motility, growth and development of H. contortus. We also studied the effect of the 'hit' compound on mitochondrial function by measuring oxygen consumption. Among the 400 Pathogen Box compounds, we identified one chemical, called tolfenpyrad (compound identification code: MMV688934) that reproducibly inhibits xL3 motility as well as L4 motility, growth and development, with IC50 values ranging between 0.02 and 3 μM. An assessment of mitochondrial function showed that xL3s treated with tolfenpyrad consumed significantly less oxygen than untreated xL3s, which was consistent with specific inhibition of complex I of the respiratory electron transport chain in arthropods. Given that tolfenpyrad was developed as a pesticide and has already been tested for absorption, distribution, excretion, biotransformation, toxicity and metabolism, it shows considerable promise for hit-to-lead optimisation and/or repurposing for use against H. contortus and other parasitic nematodes. Future work should assess its activity against hookworms and other pathogens that cause neglected tropical diseases.

Dirichlet process mixture models with pairwise constraints for data clustering

The Dirichlet process mixture (DPM) model, a typical Bayesian nonparametric model, can infer the number of clusters automatically, and thus performing priority in data clustering. This paper investigates the influence of pairwise constraints in the DPM model. The pairwise constraint, known as two types: must-link (ML) and cannot-link (CL) constraints, indicates the relationship between two data points. We have proposed two relevant models which incorporate pairwise constraints: the constrained DPM (C-DPM) and the constrained DPM with selected constraints (SC-DPM). In C-DPM, the concept of chunklet is introduced. ML constraints are compiled into chunklets and CL constraints exist between chunklets. We derive the Gibbs sampling of the C-DPM based on chunklets. We further propose a principled approach to select the most useful constraints, which will be incorporated into the SC-DPM. We evaluate the proposed models based on three real datasets: 20 Newsgroups dataset, NUS-WIDE image dataset and Facebook comments datasets we collected by ourselves. Our SC-DPM performs priority in data clustering. In addition, our SC-DPM can be potentially used for short-text clustering.