[London] : [publisher not identified], [2019]

Emanuel Ben-David, Bala Rajaratnam.

Source: Electronic Journal of Statistics, Volume 14, Number 1, 580--604.

Abstract:
The Wishart distribution defined on the open cone of positive-definite matrices plays a central role in multivariate analysis and multivariate distribution theory. Its domain of parameters is often referred to as the Gindikin set. In recent years, varieties of useful extensions of the Wishart distribution have been proposed in the literature for the purposes of studying Markov random fields and graphical models. In particular, generalizations of the Wishart distribution, referred to as Type I and Type II (graphical) Wishart distributions introduced by Letac and Massam in Annals of Statistics (2007) play important roles in both frequentist and Bayesian inference for Gaussian graphical models. These distributions have been especially useful in high-dimensional settings due to the flexibility offered by their multiple-shape parameters. Concerning Type I and Type II Wishart distributions, a conjecture of Letac and Massam concerns the domain of multiple-shape parameters of these distributions. The conjecture also has implications for the existence of Bayes estimators corresponding to these high dimensional priors. The conjecture, which was first posed in the Annals of Statistics, has now been an open problem for about 10 years. In this paper, we give a necessary condition for the Letac and Massam conjecture to hold. More precisely, we prove that if the Letac and Massam conjecture holds on a decomposable graph, then no two separators of the graph can be nested within each other. For this, we analyze Type I and Type II Wishart distributions on appropriate Markov equivalent perfect DAG models and succeed in deriving the aforementioned necessary condition. This condition in particular identifies a class of counterexamples to the conjecture.

We present a probabilistic framework for studying adversarial attacks on discrete data. Based on this framework, we derive a perturbation-based method, Greedy Attack, and a scalable learning-based method, Gumbel Attack, that illustrate various tradeoffs in the design of attacks. We demonstrate the effectiveness of these methods using both quantitative metrics and human evaluation on various state-of-the-art models for text classification, including a word-based CNN, a character-based CNN and an LSTM. As an example of our results, we show that the accuracy of character-based convolutional networks drops to the level of random selection by modifying only five characters through Greedy Attack.

Model Stealing (MS) attacks allow an adversary with black-box access to a Machine Learning model to replicate its functionality, compromising the confidentiality of the model. Such attacks train a clone model by using the predictions of the target model for different inputs. The effectiveness of such attacks relies heavily on the availability of data necessary to query the target model. Existing attacks either assume partial access to the dataset of the target model or availability of an alternate dataset with semantic similarities.

This paper proposes MAZE -- a data-free model stealing attack using zeroth-order gradient estimation. In contrast to prior works, MAZE does not require any data and instead creates synthetic data using a generative model. Inspired by recent works in data-free Knowledge Distillation (KD), we train the generative model using a disagreement objective to produce inputs that maximize disagreement between the clone and the target model. However, unlike the white-box setting of KD, where the gradient information is available, training a generator for model stealing requires performing black-box optimization, as it involves accessing the target model under attack. MAZE relies on zeroth-order gradient estimation to perform this optimization and enables a highly accurate MS attack.

Our evaluation with four datasets shows that MAZE provides a normalized clone accuracy in the range of 0.91x to 0.99x, and outperforms even the recent attacks that rely on partial data (JBDA, clone accuracy 0.13x to 0.69x) and surrogate data (KnockoffNets, clone accuracy 0.52x to 0.97x). We also study an extension of MAZE in the partial-data setting and develop MAZE-PD, which generates synthetic data closer to the target distribution. MAZE-PD further improves the clone accuracy (0.97x to 1.0x) and reduces the query required for the attack by 2x-24x.

David Corlin Marchand, Ioan Manolescu.

Source: Bernoulli, Volume 26, Number 3, 1665--1705.

Abstract:
We study randomly growing trees governed by the affine preferential attachment rule. Starting with a seed tree $S$, vertices are attached one by one, each linked by an edge to a random vertex of the current tree, chosen with a probability proportional to an affine function of its degree. This yields a one-parameter family of preferential attachment trees $(T_{n}^{S})_{ngeq |S|}$, of which the linear model is a particular case. Depending on the choice of the parameter, the power-laws governing the degrees in $T_{n}^{S}$ have different exponents. We study the problem of the asymptotic influence of the seed $S$ on the law of $T_{n}^{S}$. We show that, for any two distinct seeds $S$ and $S'$, the laws of $T_{n}^{S}$ and $T_{n}^{S'}$ remain at uniformly positive total-variation distance as $n$ increases. This is a continuation of Curien et al. ( J. Éc. Polytech. Math. 2 (2015) 1–34), which in turn was inspired by a conjecture of Bubeck et al. ( IEEE Trans. Netw. Sci. Eng. 2 (2015) 30–39). The technique developed here is more robust than previous ones and is likely to help in the study of more general attachment mechanisms.

Christian Hirsch, Christian Mönch.

Source: Bernoulli, Volume 26, Number 2, 927--947.

Abstract:
This paper considers two asymptotic properties of a spatial preferential-attachment model introduced by E. Jacob and P. Mörters (In Algorithms and Models for the Web Graph (2013) 14–25 Springer). First, in a regime of strong linear reinforcement, we show that typical distances are at most of doubly-logarithmic order. Second, we derive a large deviation principle for the empirical neighbourhood structure and express the rate function as solution to an entropy minimisation problem in the space of stationary marked point processes.

Eduard Belitser, Nurzhan Nurushev.

Source: Bernoulli, Volume 26, Number 1, 191--225.

Abstract:
In the general signal$+$noise (allowing non-normal, non-independent observations) model, we construct an empirical Bayes posterior which we then use for uncertainty quantification for the unknown, possibly sparse, signal. We introduce a novel excessive bias restriction (EBR) condition, which gives rise to a new slicing of the entire space that is suitable for uncertainty quantification. Under EBR and some mild exchangeable exponential moment condition on the noise, we establish the local (oracle) optimality of the proposed confidence ball. Without EBR, we propose another confidence ball of full coverage, but its radius contains an additional $sigma n^{1/4}$-term. In passing, we also get the local optimal results for estimation , posterior contraction problems, and the problem of weak recovery of sparsity structure . Adaptive minimax results (also for the estimation and posterior contraction problems) over various sparsity classes follow from our local results.

Elisa Filevich
Jan 21, 2015; 35:1082-1088
BehavioralSystemsCognitive

Italian translation of the BIS press release about the BIS Quarterly Review, December 2018

The UK's plans to launch a smartphone application to track potential COVID-19 infections won't include Apple and Google. The country's National Health Service has designed its own mobile software to do contact tracing of people exposed to the coronavirus. The NHS reportedly found that its own tech works "sufficiently well." The NHS chose a centralized model for its data collection and storage.

Contact tracing is a big job, like trying to drain an ocean with a teaspoon. It involves finding people who have been exposed to the coronavirus and testing them to determine if they are infected or are carriers. Public health officials then can take necessary steps to prevent the virus' spread. It's a perfect fit for CRM, and Salesforce's core technology is coming to the forefront.

Awareness for surprising sensory events is shaped by prior belief inferred from past experience. Here, we combined hierarchical Bayesian modeling with fMRI on an associative learning task in 28 male human participants to characterize the effect of the prior belief of tactile events on connections mediating the outcome of perceptual decisions. Activity in anterior insular cortex (AIC), premotor cortex (PMd), and inferior parietal lobule (IPL) were modulated by prior belief on unexpected targets compared with expected targets. On expected targets, prior belief decreased the connection strength from AIC to IPL, whereas it increased the connection strength from AIC to PMd when targets were unexpected. Individual differences in the modulatory strength of prior belief on insular projections correlated with the precision that increases the influence of prediction errors on belief updating. These results suggest complementary effects of prior belief on insular-frontoparietal projections mediating the precision of prediction during probabilistic tactile learning.

SIGNIFICANCE STATEMENT In a probabilistic environment, the prior belief of sensory events can be inferred from past experiences. How this prior belief modulates effective brain connectivity for updating expectations for future decision-making remains unexplored. Combining hierarchical Bayesian modeling with fMRI, we show that during tactile associative learning, prior expectations modulate connections originating in the anterior insula cortex and targeting salience-related and attention-related frontoparietal areas (i.e., parietal and premotor cortex). These connections seem to be involved in updating evidence based on the precision of ascending inputs to guide future decision-making.

Bombing ground targets from the air is tricky and not always accurate. But a new type of bomb creates an unimaginable level of destruction

The USS Barb had an unusual target in its sights in 1945 - one that wasn't even in the water. It was a Japanese supply train on the island of Karafuto

Contact tracing is a big job, like trying to drain an ocean with a teaspoon. It involves finding people who have been exposed to the coronavirus and testing them to determine if they are infected or are carriers. Public health officials then can take necessary steps to prevent the virus' spread. It's a perfect fit for CRM, and Salesforce's core technology is coming to the forefront.

The chief of Millbrook First Nation says the Nova Scotia RCMP's competition for its new communication centre was "distorted and biased," arbitrarily inflating the price of his community's proposal to reach a "predetermined outcome."

Jamal Campbell was never quite sure how the box of food ended up at his home while growing up in the community of Jane and Finch in Toronto, but he understood what it meant to his family.

Ontario Premier Doug Ford says the Toronto Maple Leafs' parent company has been in contact with the province about the possibility of Canada's biggest city serving as a so-called "hockey pod" for teams should the NHL resume its season.

In what could herald a course reversal for the UK's National Health Service, health officials in that country have reportedly asked a team of developers to "investigate" switching its contact tracing app to a cross-platform API provided by Apple and Google.

Several European countries, including Switzerland and Germany, are demanding all user data generated by coronavirus contact tracing apps be stored on-device, rather than aggregated on a centralized server.

Apple is supposedly developing new mental health features for the Apple Watch line that allow the wearable to detect panic attacks, with the capability potentially slated for launch with a next-generation "watchOS 7" this fall.

THE limit for in-store card transactions by contactless-enabled cards is to increase soon.

THIS time last year the Scottish Government, Police Scotland and the Crown Office were united in great fanfare about groundbreaking new legislation designed to tackle domestic abuse.

FRANKLY? The restaurant world is going mad. Consider this: I order a home delivery from Yu-ca-taco early on Friday evening. For Saturday night. By text of course.

Two football players from John Jay High School in San Antonio, Texas, could be facing criminal charges after appearing to intentionally tackle a referee during a game on Friday night.

Stronger data practices can help leaders better utilize data as a way to deeply understand the students they serve.

For three decades now, the artist and playwright John Byrne has been sitting regularly for photographer David Eustace, the Glasgow-born photographer who left school at 16 and joined first the navy and then the prison service before settling on a career behind a camera.

TactX is the 3D coaching tool that can revolutionise the work of every coach and you can take advantage of a number of exciting new features from just €5 a month.

Source: El Periódico - Las políticas climáticas que se centran en la agricultura y los bosques podrían llevar al aumento de los precios de los alimentos, pero reducir la deforestación e incrementar la captura de carbono en la agricultura podría reducir significativamente las emisiones de gases de efecto invernadero, evitando riesgos para la seguridad alimentaria, según un nuevo estudio publicado en 'Environmental Research Letters'.

For mothers of children with chronic medical conditions or disabilities, such as epilepsy or cerebral palsy, a history of maternal unresolved grief regarding the child's diagnosis has been associated with insecure infant-mother attachment.

Unresolved grief related to a preterm birth is associated with the development of insecure infantmother attachment. Mothers with resolved grief after preterm birth are 2.9 times as likely to have securely attached infants, compared with mothers with unresolved grief. (Read the full article)

Preterm infants from 33 gestational weeks can detect specific shape features (prism and cylinder) by touch (without visual control), and remember them; however, nothing is known about such abilities earlier in development.

The preterm infant, even when very immature (from 28 gestational weeks), is endowed with tactile sensory abilities: the exploration and memorization of an object by touch, the discrimination of a novel object, and the recognition of a familiar object after interference. (Read the full article)

Siblings and other contacts of abused children, especially twins, are thought to be at higher risk for abuse than other children. However, the rate at which screening tests identify injuries in contacts is currently unknown.

Contacts of abused children with serious injuries have fractures identified on skeletal survey at significant rates. Twins are at substantially increased risk for fracture. Physical examination findings were not sensitive for fractures. (Read the full article)

Respiratory infectious symptoms are common during the first year of life. Day care attendance, older siblings, and lack of breastfeeding have been considered as possible factors influencing early respiratory tract infections.

Children with early dog contacts seem to have fewer infectious respiratory symptoms and diseases, especially otitis, during the first year of life. (Read the full article)

Studies have shown dysfunction in the baroreflex mechanism and the autonomic nervous system, particularly in the sympathetic nervous system, in the pathophysiology of chronic fatigue syndrome, postural orthostatic tachycardia syndrome, and syncope.

Vitamin B₁₂ deficiency is associated with postural orthostatic tachycardia syndrome in adolescence. (Read the full article)

The Government Accountability Office finds that parents often have a hard time initiating complaints about special education services, but that these barriers don't affect all parents in the same way.

We analyse how Latvia secured a point against Germany at UEFA EURO 2004.

Catch your opponents cold by improving your team's counterattack with this training exercise.

The justices hear arguments Nov. 12 on the Trump administration's effort to end deportation relief under Deferred Action for Childhood Arrivals, in a case pitting the administration and GOP-leaning states against a host of education and advocacy groups.

La rechute est le risque majeur des pays post-conflit et l'une des principales raisons de cette rechute s'appelle la corruption.

The Government Accountability Office finds that parents often have a hard time initiating complaints about special education services, but that these barriers don't affect all parents in the same way.

Rangers vice-chairman John Bennett has criticised the SPFL for their 'cynical tactics' over their labelling of the club's dossier as a "smoking gun" - and insists the issue is not merely Gers vs the governing body.

Students enrolled in a crisis management course are examining the pandemic and its lessons and developing recommendations that the government and even individuals can follow. It's the kind of real-world experience that students can expect to have at Penn State Abington.

Faculty and students at Penn State College of Medicine and Penn State College of Nursing have launched a Contact Tracing Task Force in an effort to mitigate the spread of COVID-19.

In this first extract from the new UEFA EURO 2016 technical report, the expert panel reflect on the decreasing proportion of goals scored from counterattacks.

Connected storage is more versatile storage, but that's just one of the many benefits of buying a NAS device. We've tested the top models to tell you which one will work best for your home or small business.

Source: www.youtube.com - Thursday, May 07, 2020

After two unprofitable forays into augmented reality hardware, Snap is doubling down on its social camera eyewear with the $380 Spectacles 3 glasses featuring more 3D capabilities.