We consider Bayesian inference when only a limited number of noisy log-likelihood evaluations can be obtained. This occurs for example when complex simulator-based statistical models are fitted to data, and synthetic likelihood (SL) is used to form the noisy log-likelihood estimates using computationally costly forward simulations. We frame the inference task as a Bayesian sequential design problem, where the log-likelihood function is modelled with a hierarchical Gaussian process (GP) surrogate model, which is used to efficiently select additional log-likelihood evaluation locations. Motivated by recent progress in batch Bayesian optimisation, we develop various batch-sequential strategies where multiple simulations are adaptively selected to minimise either the expected or median loss function measuring the uncertainty in the resulting posterior. We analyse the properties of the resulting method theoretically and empirically. Experiments with toy problems and three simulation models suggest that our method is robust, highly parallelisable, and sample-efficient.

Many real world tasks exhibit rich structure that is repeated across different parts of the state space or in time. In this work we study the possibility of leveraging such repeated structure to speed up and regularize learning. We start from the KL regularized expected reward objective which introduces an additional component, a default policy. Instead of relying on a fixed default policy, we learn it from data. But crucially, we restrict the amount of information the default policy receives, forcing it to learn reusable behaviours that help the policy learn faster. We formalize this strategy and discuss connections to information bottleneck approaches and to the variational EM algorithm. We present empirical results in both discrete and continuous action domains and demonstrate that, for certain tasks, learning a default policy alongside the policy can significantly speed up and improve learning.

Scientific Computing relies on executing computer algorithms coded in some programming languages. Given a particular available hardware, algorithms speed is a crucial factor. There are many scientific computing environments used to code such algorithms. Matlab is one of the most tremendously successful and widespread scientific computing environments that is rich of toolboxes, libraries, and data visualization tools. OpenCV is a (C++)-based library written primarily for Computer Vision and its related areas. This paper presents a comparative study using 20 different real datasets to compare the speed of Matlab and OpenCV for some Machine Learning algorithms. Although Matlab is more convenient in developing and data presentation, OpenCV is much faster in execution, where the speed ratio reaches more than 80 in some cases. The best of two worlds can be achieved by exploring using Matlab or similar environments to select the most successful algorithm; then, implementing the selected algorithm using OpenCV or similar environments to gain a speed factor.

We revisit residual algorithms in both model-free and model-based reinforcement learning settings. We propose the bidirectional target network technique to stabilize residual algorithms, yielding a residual version of DDPG that significantly outperforms vanilla DDPG in the DeepMind Control Suite benchmark. Moreover, we find the residual algorithm an effective approach to the distribution mismatch problem in model-based planning. Compared with the existing TD($k$) method, our residual-based method makes weaker assumptions about the model and yields a greater performance boost.

We study the transfer of adversarial robustness of deep neural networks between different perturbation types. While most work on adversarial examples has focused on $L_\infty$ and $L_2$-bounded perturbations, these do not capture all types of perturbations available to an adversary. The present work evaluates 32 attacks of 5 different types against models adversarially trained on a 100-class subset of ImageNet. Our empirical results suggest that evaluating on a wide range of perturbation sizes is necessary to understand whether adversarial robustness transfers between perturbation types. We further demonstrate that robustness against one perturbation type may not always imply and may sometimes hurt robustness against other perturbation types. In light of these results, we recommend evaluation of adversarial defenses take place on a diverse range of perturbation types and sizes.

Amazon's AI assistant will soon be able to speak more than just English in the U.S. The internet giant has launched a new program that will allow U.S. developers to build skills for Spanish-speaking users. In doing so, it will allow Amazon to build a more robust experience for Alexa users when it launches full Spanish-language support later this year. Amazon's AI assistant will soon be able to speak more than just English. 'We're excited to announce that now developers can start building skills for Spanish-speaking customers in the US using the Alexa Skills Kit with the new Spanish for US voice model,' the company wrote in a blog post on Monday. 'Skills that developers create now and are certified for publication will be available for participants in the Alexa Preview program, and to all customers when Alexa launches in the US with Spanish language support later this year.'

NASA's InSight lander has captured stunning images of sunrise and sunset from the Martian surface. Stunning images taken from the Insight's robotic arm shows the Red Planet landscape and the sun rising and setting at the equivalent of 5:30 am and 6:30 pm Mars time. The sun looks especially small because it's further away from Mars than the Earth so it is about two-thirds the size as seen on our planet. The Insight lander has been busy recently; just a handful of months into its exploratory mission on Mars it made a groundbreaking discovery after it detected for the first time a quake on Mars, or a'Marsquake'. NASA's stationary InSight lander has captured what sunrise and sunset looks like on Mars.

Amazon said that by the end of the next decade packages in the company's warehouses could be readied for delivery without touching a single human hand. In a report from Reuters, Director of Amazon Robotics Fulfillment, Scott Anderson, told reporters that there is still plenty of progress to be made before robots take over its warehouses. 'In the current form, the technology is very limited. The technology is very far from the fully automated workstation that we would need,' Anderson told Reuters in a walk through of one of its facilities in Baltimore. Robots still have a long way to go before the replace human hands in Amazon's factories said one of the company's executives in a recent walk through of a facility in Baltimore.

President Trump began his trade war with China out of concern that Beijing was using unfair economic practices to prevent the United States from dominating next-generation technologies like autonomous vehicles, advanced telecommunications and artificial intelligence. But as the two countries move closer to a trade deal, it seems increasingly unlikely that China will give ground in a crucial area that could determine which country wins the technology race. Despite months of pressure from the White House, Chinese negotiators have so far refused to relax tight regulations that block multinational companies from moving data they gather on their Chinese customers' purchases, habits and whereabouts out of the country. Such data is crucial as industries build next-generation technologies. With another round of trade talks underway this week, the United States and China appear headed toward an agreement that could end the monthslong trade war and lift tariffs on hundreds of billions of dollars of products.

Rapper, producer and technology entrepreneur will.i.am "The investment that society has put in AI (artificial intelligence) surpasses the investment for HI, which is human intelligence," said Will.i.am during a panel at the Dell Technologies World 2019 Conference on Wednesday. "It's so lopsided that subconsciously we know that we haven't invested in our youth, in our communities. We haven't invested in humanity to keep up with intelligent machines." Will.i.am was joined by Dell Chief Marketing Officer Allison Dew as well as Brynn Putnam, founder and CEO of Mirror, which makes a full-length mirror that doubles as a screen for streaming home workout classes.