China has already taken a significant step into Germany. In the Rheinhausen district of Duisburg, trains are now rolling across the site where steelworkers once fought unsuccessfully to save their mill in 1987 while shipyard cranes stack up containers on the banks of the Rhine River. This is the precise point where the New Silk Road, China's massive infrastructure project, comes to an end. The site in Duisburg is known as Logport I and it is one of the largest container ports in Europe. Twenty-five trains arrive each week at Terminal DIT, also known as the China Terminal, after having traveled the more than 10,000 kilometers from Chongqing across Kazakhstan, Russia, Belarus and Poland. Four years ago, Chinese President Xi Jinping visited the inland port.

The AI Expo Europe, the leading artificial intelligence Conference & Exhibition event will take place on 27-28th June at Amsterdam's RAI. It is a showcase of next generation technologies and strategies from the world of Artificial Intelligence, an opportunity to explore and discover the practical and successful implementation of AI in driving forward your business in 2018 and beyond. Our high-level conferences will bring together forward thinking brands, market leaders, AI evangelists and hot start-ups to explore and debate the advancements in Artificial Intelligence and the impacts within the Enterprise & Consumer sectors as well as Development platforms and Digital Transformation opportunities. Topics covered include Business Intelligence, Deep Learning, Machine Learning, AI Algorithms, Data & Analytics, Virtual Assistants & Chatbots as well as case study based presentations proving an insight into the deployment of AI across different verticals. The AI Expo will bring together over 1,000 visitors over the two days including IT decision makers, developers & designers, heads of innovation, brand managers, data analysts and scientists, start-ups and innovators, tech providers, c-level executives and venture capitalists. The AI Expo will be co-hosted alongside the IoT Tech Expo, the largest global gathering for the Internet of Things sector, and Blockchain Expo, with both shows attracting in excess of 1,000 attendees for two days of insightful content covering the whole ecosystem surrounding AI, IoT & Blockchain.

Prof. Dr. Edouard Hannezo, Assistant Professor at the Institute of Science and Technology (IST) Austria, is interested in understanding how cells "know" how to make the right decisions at the right time and at the right place during development and normal tissue homeostasis, as well as how these decisions are dysregulated during cancer initiation.

A Japanese construction company recently debuted two designs for robot workers that could make up for the growing lack of human construction workers. In a report from The Daily Mail, these construction robots will only be working during evening hours. One of the robots demonstrated by Shimizu Corp. is already being used in several Japanese construction sites. Called Robo-Buddy, the automaton lifted a bunch of wooden boards before hauling them to the nearest elevator. The Robo-Buddy and its partner, the Robo-Welder, featured robotic arms that can twist and turn to fit in various spaces. Shimizu expected to start deploying them en masse in the latter half of 2018.

The European Robotics Forum 2018 (ERF2018), the most influential meeting of the robotics community in Europe, took place in Tampere on 13-15 March 2018. ERF2018 brought together over 900 leading scientists, companies, and policymakers. Under the theme "Robots and Us", the over 50 workshops cover current societal and technical themes, including human-robot-collaboration and how robotics can improve industrial productivity and service sector operations.

With strawberry picking season well under way - but migrant labour in short supply in several countries - we look at the various robots being developed around the world to help producers harvest this most popular fruit. Next time you buy strawberries take a look a good look in the punnet. Do the berries still have the stem attached or has it been plucked off leaving only the green hat of leaves called the calyx? You may not think that matters, but it's a key consideration for growers as they contemplate the merits of a range of robotic prototypes that promise to pick strawberries as fast and as carefully as humans. Whether the berry is plucked or whether the stalk is snipped through and kept attached is one critical difference between the concepts that Spanish, Belgian, British and US engineers are testing, ready to roll out in fields as soon as next year.

The Sensor, Observation, Sample, and Actuator (SOSA) ontology provides a formal but lightweight general-purpose specification for modeling the interaction between the entities involved in the acts of observation, actuation, and sampling. SOSA is the result of rethinking the W3C-XG Semantic Sensor Network (SSN) ontology based on changes in scope and target audience, technical developments, and lessons learned over the past years. SOSA also acts as a replacement of SSN's Stimulus Sensor Observation (SSO) core. It has been developed by the first joint working group of the Open Geospatial Consortium (OGC) and the World Wide Web Consortium (W3C) on Spatial Data on the Web. In this work, we motivate the need for SOSA, provide an overview of the main classes and properties, and briefly discuss its integration with the new release of the SSN ontology as well as various other alignments to specifications such as OGC's Observations and Measurements (O&M), Dolce-Ultralite (DUL), and other prominent ontologies. We will also touch upon common modeling problems and application areas related to publishing and searching observation, sampling, and actuation data on the Web. The SOSA ontology and standard can be accessed at https://www.w3.org/TR/vocab-ssn/. Keywords: Ontology, Sensor, Observation, Actuator, Linked Data, Web of Things, Internet of Things, Schema.org 1. Introduction and Motivation In their broadest definition sensors detect and react to changes in the environment that directly or indirectly reveal the value of a property. The process of determining this, not necessarily numeric, value is called an observation.

Policy evaluation, commonly referred to as value function approximation, is an important and central part in many reinforcement learning (RL) algorithms [27], whose task is to estimate value functions for a fixed policy in a discounted Markov Decision Process (MDP) environment. The value function of each state specifies the accumulated reward an agent would receive in the future by following the fixed policy from that state. Value functions have been widely investigated in RL applications, and it can provide insightful and important information for the agent to obtain an optimal policy, such as important board configurations in Go [24], failure probabilities of large telecommunication networks [9], taxi-out times at large airports [2] and so on. Despite the value functions can be approximated by different ways, the simplest form, linear approximations, are still widely adopted and studied due to their good generalization abilities, relatively efficient computation and solid theoretical guarantees[27, 7, 13, 16]. Temporal Difference (TD) learning is a common approach to this policy evaluation with linear function approximation problem[27]. These typical TD algorithms can be divided into two categories: gradient based methods (e.g., GTD(λ) [28]) and least-square (LS) based methods (e.g., LSTD(λ)[4]). A good survey on these algorithms can be found in [17, 6, 12, 7, 13]. 1 As the development of information technologies, high-dimensional data is widely seen in RL applications [26, 30, 23], which brings serious challenges to design scalable and computationally efficient algorithms for the linear value function approximation problem. To address this practical issue, several approaches have been developed for efficient and effective value function approximation.

Applying reinforcement learning in physical-world tasks is extremely challenging. It is commonly infeasible to sample a large number of trials, as required by current reinforcement learning methods, in a physical environment. This paper reports our project on using reinforcement learning for better commodity search in Taobao, one of the largest online retail platforms and meanwhile a physical environment with a high sampling cost. Instead of training reinforcement learning in Taobao directly, we present our approach: first we build Virtual Taobao, a simulator learned from historical customer behavior data through the proposed GAN-SD (GAN for Simulating Distributions) and MAIL (multi-agent adversarial imitation learning), and then we train policies in Virtual Taobao with no physical costs in which ANC (Action Norm Constraint) strategy is proposed to reduce over-fitting. In experiments, Virtual Taobao is trained from hundreds of millions of customers' records, and its properties are compared with the real environment. The results disclose that Virtual Taobao faithfully recovers important properties of the real environment. We also show that the policies trained in Virtual Taobao can have significantly superior online performance to the traditional supervised approaches. We hope our work could shed some light on reinforcement learning applications in complex physical environments.

Although neural machine translation (NMT) with the encoder-decoder framework has achieved great success in recent times, it still suffers from some drawbacks: RNNs tend to forget old information which is often useful in the current step and the encoder only operates over words without considering word relationship. To solve these problems, we introduce relation networks (RNs) to learn better representations of the source. In our method RNs are used to associate source words with each other so that the source representation can memorize all the source words and also contain the relationship between them. Then the source representations and all the relations are fed into the attention component together while decoding, with the main encoder-decoder architecture unchanged. Experiments on several data sets show that our method can improve the translation performance significantly over the conventional encoder-decoder model, and can even outperform the approach involving supervised syntactic knowledge.