Representations of probability measures in reproducing kernel Hilbert spaces provide a flexible framework for fully nonparametric hypothesis tests of independence, which can capture any type of departure from independence, including nonlinear associations and multivariate interactions. However, these approaches come with an at least quadratic computational cost in the number of observations, which can be prohibitive in many applications. Arguably, it is exactly in such large-scale datasets that capturing any type of dependence is of interest, so striking a favourable tradeoff between computational efficiency and test performance for kernel independence tests would have a direct impact on their applicability in practice. In this contribution, we provide an extensive study of the use of large-scale kernel approximations in the context of independence testing, contrasting block-based, Nystrom and random Fourier feature approaches. Through a variety of synthetic data experiments, it is demonstrated that our novel large scale methods give comparable performance with existing methods whilst using significantly less computation time and memory.

This paper focuses on the problem of explaining predictions of psychological attributes such as attractiveness, happiness, confidence and intelligence from face photographs using deep neural networks. Since psychological attribute datasets typically suffer from small sample sizes, we apply transfer learning with two base models to avoid overfitting. These models were trained on an age and gender prediction task, respectively. Using a novel explanation method we extract heatmaps that highlight the parts of the image most responsible for the prediction. We further observe that the explanation method provides important insights into the nature of features of the base model, which allow one to assess the aptitude of the base model for a given transfer learning task. Finally, we observe that the multiclass model is more feature rich than its binary counterpart. The experimental evaluation is performed on the 2222 images from the 10k US faces dataset containing psychological attribute labels as well as on a subset of KDEF images.

We consider a group of Bayesian agents who try to estimate a state of the world $\theta$ through interaction on a social network. Each agent $v$ initially receives a private measurement of $\theta$: a number $S_v$ picked from a Gaussian distribution with mean $\theta$ and standard deviation one. Then, in each discrete time iteration, each reveals its estimate of $\theta$ to its neighbors, and, observing its neighbors' actions, updates its belief using Bayes' Law. This process aggregates information efficiently, in the sense that all the agents converge to the belief that they would have, had they access to all the private measurements. We show that this process is computationally efficient, so that each agent's calculation can be easily carried out. We also show that on any graph the process converges after at most $2N \cdot D$ steps, where $N$ is the number of agents and $D$ is the diameter of the network. Finally, we show that on trees and on distance transitive-graphs the process converges after $D$ steps, and that it preserves privacy, so that agents learn very little about the private signal of most other agents, despite the efficient aggregation of information. Our results extend those in an unpublished manuscript of the first and last authors.

It is the future of shipping - and there's not a single sailor on board. Rolls Royce has revealed planed for fleets of'drone ships' to ferry carry around the world - all controlled from a central'holodeck'. It believes an entirely unmanned ship could take to the seas by 2020. Rolls Royce said it has already begun testing the technology needed to make the ships a reality, and expected them to take to the sea by the end of the decade. Cameras would beam 360-degree views from the drone ship back to operators based in a virtual bridge.

The Economist recently wrote a moving paean to the fate of autistic people in the western world. It revealed that one in 68 people in America are believed to be autistic, but western society appears to be doing a terrible job on actually helping the autistic among us contribute and feel valued. Although around half of those with autism are of average intelligence or above, they do far worse than they should at school and at work. In France, almost 90% of autistic children attend primary school, but only 1% make it to high school. Figures from America, which works harder to include autistic pupils, suggest that less than half graduate from high school.

Starting this fall when iOS 10 and macOS Sierra are launched, there will be (optional) changes to Apple's privacy policy. The company has finally acknowledged publicly that it needs to collect at least some types of information if it wants to advance with its artificial intelligence ambitions. The change was announced by Apple SVP Craig Federighi, who said that the company will collect information in a different way to before, as it seeks to improve the ability of Siri and the iPhone to predict the information the user wants. Federighi touted this approach as differential privacy. Wired published an extensive piece about what differential privacy means and how Apple plans to implement it.

It's time for another edition of "What TNW is reading." If you're not familiar with them, take a moment and enjoy the previous edition about the US election. After having first been colonized by the Spanish, the country – in its own weird way – is part of the Kingdom of the Netherlands and one of the busiest ports in the Northern Hemisphere. One of its four official languages is English and they managed to smush them all together to create Papiamentu, the local language. Speaking all four languages fluently, Cecil is currently adding French and Russian to the mix… he's such an?????????.

A tipster identified Green as the shooter several days after the killing, according to the Post. When detectives learned Green was on house arrest, they asked the city Pretrial Services Agency for his whereabouts on the day of the shooting. The GPS tracker showed he was at home, police said -- and remained there days later. Investigators checked camera footage from the area near the shooting and spotted a gunman with an obvious limp, according to FoxDC. Inside Green's home, police with a search warrant discovered a box in his living room containing the prosthetic leg with the tracking device still attached to the ankle.

While it may look like a natural phenomenon, this watery sinkhole is anything but. Named Covão do Conchos it is regarded as one of Portugal's top secret attractions. The mesmerising whirlpool, which has been caught in high definition by a circling drone, is the work of human engineering and can be found in an otherwise tranquil lake in Serra da Estrela Natural Park. People say the opening in Portugal's largest protected area looks like it's from an otherworldly dimension. However, it is far less mystical and far more fascinating.

The Californian desert rushes up in front of me. I can see the runway at Edwards Air Force Base emerging clearly from the hills, and I try to keep the nose of my spacecraft pointed straight down the centre. I am flying the Dream Chaser space plane back from a stint at the International Space Station (ISS), and am keenly aware of my delicate cargo – and the craft's past failures. I'm seated in front of three computer monitors, which show my view out of the cockpit, and rear and side views of the spacecraft as it descends. To go easy on me, the Draper crew starts the simulation after the Dream Chaser has already entered Earth's atmosphere and headed down towards the ground, so all I have to do is aim it straight at the runway.