We provide simple schemes to build Bayesian Neural Networks (BNNs), block by block, inspired by a recent idea of computation skeletons. We show how by adjusting the types of blocks that are used within the computation skeleton, we can identify interesting relationships with Deep Gaussian Processes (DGPs), deep kernel learning (DKL), random features type approximation and other topics. We give strategies to approximate the posterior via doubly stochastic variational inference for such models which yield uncertainty estimates. We give a detailed theoretical analysis and point out extensions that may be of independent interest. As a special case, we instantiate our procedure to define a Bayesian {\em additive} Neural network -- a promising strategy to identify statistical interactions and has direct benefits for obtaining interpretable models.

Future predictions on sequence data (e.g., videos or audios) require the algorithms to capture non-Markovian and compositional properties of high-level semantics. Context-free grammars are natural choices to capture such properties, but traditional grammar parsers (e.g., Earley parser) only take symbolic sentences as inputs. In this paper, we generalize the Earley parser to parse sequence data which is neither segmented nor labeled. This generalized Earley parser integrates a grammar parser with a classifier to find the optimal segmentation and labels, and makes top-down future predictions. Experiments show that our method significantly outperforms other approaches for future human activity prediction.

With the world moving towards being increasingly dependent on computers and automation, one of the main challenges in the current decade has been to build secure applications, systems and networks. Alongside these challenges, the number of threats is rising exponentially due to the attack surface increasing through numerous interfaces offered for each service. To alleviate the impact of these threats, researchers have proposed numerous solutions; however, current tools often fail to adapt to ever-changing architectures, associated threats and 0-days. This manuscript aims to provide researchers with a taxonomy and survey of current dataset composition and current Intrusion Detection Systems (IDS) capabilities and assets. These taxonomies and surveys aim to improve both the efficiency of IDS and the creation of datasets to build the next generation IDS as well as to reflect networks threats more accurately in future datasets. To this end, this manuscript also provides a taxonomy and survey or network threats and associated tools. The manuscript highlights that current IDS only cover 25% of our threat taxonomy, while current datasets demonstrate clear lack of real-network threats and attack representation, but rather include a large number of deprecated threats, hence limiting the accuracy of current machine learning IDS. Moreover, the taxonomies are open-sourced to allow public contributions through a Github repository.

This document gives a knowledge-oriented analysis of about 20 interesting Recognizing Textual Entailment (RTE) examples, drawn from the 2005 RTE5 competition test set. The analysis ignores shallow statistical matching techniques between T and H, and rather asks: What would it take to reasonably infer that T implies H? What world knowledge would be needed for this task? Although such knowledge-intensive techniques have not had much success in RTE evaluations, ultimately an intelligent system should be expected to know and deploy this kind of world knowledge required to perform this kind of reasoning. The selected examples are typically ones which our RTE system (called BLUE) got wrong and ones which require world knowledge to answer. In particular, the analysis covers cases where there was near-perfect lexical overlap between T and H, yet the entailment was NO, i.e., examples that most likely all current RTE systems will have got wrong. A nice example is #341 (page 26), that requires inferring from "a river floods" that "a river overflows its banks". Seems it should be easy, right? Enjoy!

Over 76,000 drones could be in the UK skies in 12 years' time -- and it could add as much as £1.1 billion (€1.3 billion) to the amount generated through farming and other primary industries. Drone technology has the potential to increase the UK's gross domestic product (GDP) by £42 billion (or 2%) by 2030, according to new research from PwC. The report, 'Skies Without Limits', estimates that more than a third of these could be used for agriculture, mining, gas and electricity. Certain drones combine normal and thermal cameras to deliver a level of insight into field crop health that is not obvious to the naked human eye. For example, thermal imaging can detect dry areas and ensure water is delivered where required.

That plastic cup you've got sitting on your desk looks pretty harmless on its own. However, add it to the rest of the plastic that humanity throws away on a daily basis and you have the makings of the estimated 5 to 13 million metric tons of plastic trash which reportedly wind up in the world's oceans every year. U.K.-based Plastic Tide founder Peter Kohler got a glimpse of the scale of this problem a decade ago -- and it changed the course of his life. "About ten years ago, I went out to the South Pacific," he told Digital Trends. "I've always been fascinated by oceans, and this was pure paradise. But it was a paradise under siege. One of the most visible ways this paradise was being besieged was with litter. It was everywhere, although we were miles from anyone. When you're sailing in the middle of nowhere, it really gets you wondering where this litter comes from and how it gets here. I came back to England and spent the next few years puzzling over how best to answer that question."

The Department of Energy pulled back the curtain on the world's most powerful supercomputer Friday. When Summit is operating at max capacity, it can run at 200 petaflops -- that's 200 quadrillion calculations per second. That smokes the previous record holder, China's Sunway TaihuLight (which has a 93 petaflop capacity). Summit is also about seven times faster than Titan, the previous US record holder which is housed at the same Oak Ridge National Lab in Tennessee. For perspective, in one hour, Summit can solve a problem that it would take a desktop computer 30 years to crack.

Making high-performance batteries for electric vehicles is a complicated, time-consuming process. So much so that engineers at Volkswagen have started using a quantum computer to simulate the chemical structures like lithium-hydrogen and carbon chains much faster. The idea is to continue using quantum computing to eventually develop a sort of blueprint for tailor-made batteries that can be optimized for different features, like weight reduction, power density or power cell assembly. "We are working hard to develop the potential of quantum computers for Volkswagen," said scientist Florian Neukart in a statement. "The simulation of electrochemical materials is an important project in this context. In this field, we are performing genuine pioneering work. We are convinced that commercially available quantum computers will open up previously unimaginable opportunities. We intend to acquire the specialist knowledge we need for this purpose now."

Photo: Robotic arm made for MIT AI lab, 1972. In 1973, the burgeoning field of artificial intelligence (AI) was at a crossroads: Governments had grown skeptical of the research and were hesitant to continue pumping millions of dollars into projects with few actionable results. So, to examine what AI had done, the United Kingdom solicited a report from a famous mathematician named Sir James Lighthill. What happened next was controlled chaos, the sort of polite brawling that only PhD's in neat suits are capable of. AI researchers erupted with anger, and possibly for good reason: It turned out that Lighthill, whose report demolished AI, did not fully understand the field he was trying to crush.

Video Friday is your weekly selection of awesome robotics videos, collected by your Automaton bloggers. We'll also be posting a weekly calendar of upcoming robotics events for the next few months; here's what we have so far (send us your events!): Let us know if you have suggestions for next week, and enjoy today's videos. Since its epic landing on Mars in 2012, rappelling down to the surface like a robot commando, the Curiosity Mars rover has been one of our favorite robots of all time, and space. Not only it's an impressive piece of engineering, it's also an amazing exploration tool to help humanity answer questions we've been asking ourselves for a very long time, including: Are we alone?