Does Interactive Conditioning Help Users Better Understand the Structure of Probabilistic Models?

Despite growing interest in probabilistic modeling approaches and availability of learning tools, people are hesitant to use them. There is a need for tools to communicate probabilistic models more intuitively and help users build, validate, use effectively or trust probabilistic models. We focus on visual representations of probabilistic models and introduce the Interactive Pair Plot (IPP) for visualization of a model's uncertainty, a scatter plot matrix of a probabilistic model allowing interactive conditioning on the model's variables. We investigate whether the use of interactive conditioning in a scatter plot matrix of a model helps users better understand variables' relations. We conducted a user study and the findings suggest that improvements in the understanding of the interaction group are the most pronounced for more exotic structures, such as hierarchical models or unfamiliar parameterizations, in comparison to the understanding of the static group. As the detail of the inferred information increases, interactive conditioning does not lead to considerably longer response times. Finally, interactive conditioning improves participants' confidence about their responses.

Efficient human-machine control with asymmetric marginal reliability input devices.

Input devices such as motor-imagery brain-computer interfaces (BCIs) are often unreliable. In theory, channel coding can be used in the human-machine loop to robustly encapsulate intention through noisy input devices but standard feedforward error correction codes cannot be practically applied. We present a practical and general probabilistic user interface for binary input devices with very high noise levels. Our approach allows any level of robustness to be achieved, regardless of noise level, where reliable feedback such as a visual display is available. In particular, we show efficient zooming interfaces based on feedback channel codes for two-class binary problems with noise levels characteristic of modalities such as motor-imagery based BCI, with accuracy <75%. We outline general principles based on separating channel, line and source coding in human-machine loop design. We develop a novel selection mechanism which can achieve arbitrarily reliable selection with a noisy two-state button. We show automatic online adaptation to changing channel statistics, and operation without precise calibration of error rates. A range of visualisations are used to construct user interfaces which implicitly code for these channels in a way that it is transparent to users. We validate our approach with a set of Monte Carlo simulations, and empirical results from a human-in-the-loop experiment showing the approach operates effectively at 50-70% of the theoretical optimum across a range of channel conditions.

Relationship between previous history of Streptococcus uberis infection and response to a challenge model.

Streptococcus uberis is the most common cause of clinical mastitis at calving in pasture-based dairy cows. Results of experimental inoculations were compared with cows' previous history of infection to help define a model for susceptibility to Str. uberis mastitis. Cows used had either no apparent history of intramammary infection (IMI) by Str. uberis or other major mastitis pathogens throughout their productive lifetime ('apparently uninfected'; AUI), or had a confirmed history of Str. uberis IMI ('historically infected'; HI). Cows were exposed to Str. uberis in sequential steps: dipping of the teat end (DIP; n=53 cows); a teat canal inoculation (TCI; n=33 cows); and, finally, intramammary inoculation challenge (IC; n=7 cows). Only cows that remained free of infection at each step progressed to the next phase. Infection rates were similar between AUI or HI cows following the DIP (9 and 17% respectively), or the TCI (75 and 68% respectively). Physical and biochemical traits of cows were examined. Analysis of traits prior to inoculations implied that HI cows produced more milk fat, while AUI cows tended to have longer teat canals. Analysis of traits for cows that became infected following DIP, implied that there was a positive association with milk fat production and negative association with somatic cell count (SCC), while there was a positive association with the duration of p.m. milking, and negative association with SCC in those cows that became infected following TCI. Only AUI cows became infected following the IC inoculation. Similarity in response to experimental inoculation between the two groups suggests that the current dip or teat canal inoculation (using a 3-mm depth of inoculation) models are not good predictors of natural resistance to Str. uberis. However, a population of cows was identified that remained uninfected after DIP, TCI and IC, and may comprise a resistant phenotype.

Streptococcus uberis-specific T cells are present in mammary gland secretions of cows and can be activated to kill S. uberis.

The presence, phenotype and function of Streptococcus uberis-specific T cells in the mammary gland secretion (MGS) and blood of cows exposed to S. uberis were assessed. MGS T cells in the udder were purified and incubated with autologous blood monocytes as antigen-presenting cells (APC). Most cows, irrespective of prior S. uberis infection status and lactation status, were shown to have S. uberis-specific T cells both in MGS and in the blood. When cells from a subgroup of cows were studied, it was found that the S. uberis-specific T cells produced high levels of interferon-gamma (IFN-γ), but low levels of interleukin-10 (IL-10). A high percentage of responding T cells were of the CD8(+) memory (CD45RO) subset. T cells from the MGS specific for S. uberis were propagated from animals during the drying off period and expanded in vitro using interleukin-2 (IL-2) and S. uberis antigens. This led to the accumulation of T cells of the CD8(+) subset bearing memory cell markers (CD45A(-), CD45RO(+)), which released high levels of IFN-γ. Four of the five T cell lines derived from the MGS of three animals had substantial direct killing activity towards S. uberis in vitro. It is concluded that there is an emergence of S. uberis-specific bactericidal T cells in the MGS of cows after infection or environmental exposure to S. uberis. Vaccines aimed at activating and expanding this T cell population in the mammary glands of cattle may offer an avenue for the prevention of mastitis caused by S. uberis.

Heifer teats sprayed in the dry period with an iodine teat sanitizer have reduced Streptococcus uberis teat-end contamination and less Streptococcus uberis intra-mammary infections at calving.

Heifers managed under pastoral conditions are at risk from Streptococcus uberis mastitis infections at calving. A total of 397 heifers from six farms around New Zealand were enrolled in a study to identify and enumerate S. uberis on teat-ends of heifers in the peri-partum period, and to understand the effect of teat-spraying in the pre-calving period on the prevalence and incidence of S. uberis mastitis post-calving. Heifers were randomly assigned to Control or Sprayed groups. Sprayed heifers were teat-sprayed once, three times a week (Monday, Wednesday and Friday) with a commercial iodine-based teat sanitizer, starting at 3 weeks prior to calving and ending at day of calving. Across three farms, all glands of cows in both groups were sampled at calving to determine S. uberis intra-mammary infection (IMI) prevalence. For all farms, clinical mastitis (CM) cases detected during the week after calving were sampled and submitted for bacteriological analysis. Swabbing of teat-ends of 54 heifers from one farm showed that heifers had a pre-existing S. uberis contamination averaging 610 colony-forming units per swab (cfu/swab), at 3 weeks prior to calving. At calving, teat-end contamination was 560 cfu/swab for Sprayed heifers and 1775 cfu/swab for Control heifers. Two weeks after calving, teat-end contamination was similar between both groups, at 30 cfu/swab. The prevalence of S. uberis IMI was significantly lower in the Sprayed (3.5% glands) vs. the Control (7.4%) heifers in the first week after calving. There was a trend for Sprayed heifers (3.6% heifers) to have a lower incidence of S. uberis CM compared with Control heifers (7.4% heifers). It is concluded that teat-spraying in the dry period is a management option that could contribute to controlling heifer S. uberis mastitis in the transition period.

Application of Streptococcus uberis multilocus sequence typing: analysis of the population structure detected among environmental and bovine isolates from New Zealand and the United Kingdom.

We recently developed a multilocus sequence typing (MLST) scheme to differentiate S. uberis isolates and facilitate an understanding of the population biology of this pathogen. The scheme was initially used to study a collection of 160 bovine milk isolates from the United Kingdom and showed that the majority of isolates were from one clonal complex (designated the ST-5 complex). Here we describe the MLST analysis of a collection of New Zealand isolates. These were obtained from diverse sources, including bovine milk, other bovine anatomical sites, and environmental sources. The complete allelic profiles of 253 isolates were determined. The collection was highly diverse and included 131 different sequence types (STs). The New Zealand and United Kingdom populations were distinct, since none of the 131 STs were represented within the previously studied collection of 160 United Kingdom S. uberis isolates. However, seven of the STs were members of the ST-5 clonal complex, the major complex within the United Kingdom collection. Two new clonal complexes were identified: ST-143 and ST-86. All three major complexes were isolated from milk, other bovine sites, and the environment. Carriage of the hasA gene, which is necessary for capsule formation, correlated with clonal complex and isolation from clinical cases of mastitis.