BBC Radio 1 has announced a major overhaul of its line-up, with the weekend schedule now starting on Fridays. Nick Grimshaw, Clara Amfo and Greg James will go down to hosting four shows a week as a result. New Friday shows have been announced for Maya Jama and The Saturdays' Mollie King as part of the changes, and Scott Mills will now host the chart show. Dev and Alice Levine will present the breakfast show from Friday to Sunday. Ben Cooper, controller of BBC Radio 1, said: "The weekend will start here at Radio 1 on a Friday morning giving our young audience that feel-good factor a day early. "It's our job at Radio 1 to reinvent the way young people listen to the radio, to disrupt traditional thinking and to look for new ways in which to grow audiences." The new schedule comes into effect in June. It's unclear how the schedule changes might affect the daytime hosts' current salaries, but a BBC spokesman said all presenters on the station are paid fairly. This marks the first time in Radio 1's history that the daytime DJs will be on a four-day week, although Chris Evans once tried to take Fridays off when he was hosting the breakfast show. In the late 1990s, he famously asked if he could host the programme only from Monday to Thursday - a request which was denied by then-controller Matthew Bannister. Jama joined the station earlier this year as the presenter of Radio 1's Greatest Hits. King has been appearing as a guest presenter with Edmondson in recent weeks but will now officially be a permanent fixture on the station. The singer, who appeared on the most recent series of Strictly Come Dancing, said: "I've grown up listening to Radio 1 and I can't wait to be one of the team.

An important problem of automated planning is validating if a plan complies with the planning domain model. Such validation is straightforward for classical sequential planning but until recently there was no such validation approach for Hierarchical Task Networks (HTN) planning. In this paper we propose a novel technique for validating HTN plans that is based on representing the HTN model as an attribute grammar and using a special parsing algorithm to verify if the plan can be generated by the grammar.

Automated planning is a major topic of research in artificial intelligence, and enjoys a long and distinguished history. The classical paradigm assumes a distinguished initial state, comprised of a set of facts, and is defined over a set of actions which change that state in one way or another. Planning in many real-world settings, however, is much more involved: an agent's knowledge is almost never simply a set of facts that are true, and actions that the agent intends to execute never operate the way they are supposed to. Thus, probabilistic planning attempts to incorporate stochastic models directly into the planning process. In this article, we briefly report on probabilistic planning through the lens of probabilistic programming: a programming paradigm that aims to ease the specification of structured probability distributions. In particular, we provide an overview of the features of two systems, HYPE and ALLEGRO, which emphasise different strengths of probabilistic programming that are particularly useful for complex modelling issues raised in probabilistic planning. Among other things, with these systems, one can instantiate planning problems with growing and shrinking state spaces, discrete and continuous probability distributions, and non-unique prior distributions in a first-order setting.

Plan recognition is  one of the fundamental problems of AI, applicable to many domains, from user interfaces to cyber security. We focus on a class of algorithms that use plan libraries as input to the recognition process. Despite the prevalence of these approaches, they lack a standard representation, and have not been compared to each other on common test bed. This paper directly addresses this gap by providing a standard plan library representation and evaluation criteria to consider. Our representation is comprehensive enough to describe a variety of known plan recognition problems, yet it can be easily applied to existing algorithms, which can be evaluated using our defined criteria. We demonstrate this technique on two known algorithms, SBR and PHATT. We provide meaningful insights both about the differences and abilities of the algorithms. We show that SBR is superior to PHATT both in terms of computation time and space, but at the expense of functionality and compact representation. We also show that depth is the single feature of a plan library that increases the complexity of the recognition, regardless of the algorithm used.

Plan- and Goal Recognition (PGR) is the task of inferring the goals and plans of an agent based on its actions. A few years ago, an approach has been introduced that successfully exploits the performance of planning systems to solve it. That way, no specialized solvers are needed and PGR benefits from present and future research in planning. The approach uses classical planning systems and needs to plan (at least) once for every possible goal. However, models in PGR are often structured in a hierarchical way, similar to Hierarchical Task Networks (HTNs). These models are strictly more expressive than those in classical planning and can describe partially ordered sets of tasks or multiple goals with interleaving plans. We present the approach PGR as HTN Planning that enables the recognition of complex agent behavior by using unmodified, off-the-shelf HTN planners. Planning is thereby needed only once, regardless of how many possible goals there are. Our evaluation shows that current planning systems are able to handle large models with thousands of possible goals and that the approach results in high recognition rates.

A key challenge in complex visuomotor control is learning abstract representations that are effective for specifying goals, planning, and generalization. To this end, we introduce universal planning networks (UPN). UPNs embed differentiable planning within a goal-directed policy. This planning computation unrolls a forward model in a latent space and infers an optimal action plan through gradient descent trajectory optimization. The plan-by-gradient-descent process and its underlying representations are learned end-to-end to directly optimize a supervised imitation learning objective. We find that the representations learned are not only effective for goal-directed visual imitation via gradient-based trajectory optimization, but can also provide a metric for specifying goals using images. The learned representations can be leveraged to specify distance-based rewards to reach new target states for model-free reinforcement learning, resulting in substantially more effective learning when solving new tasks described via image-based goals. We were able to achieve successful transfer of visuomotor planning strategies across robots with significantly different morphologies and actuation capabilities.

Recently, AAAI coordinated and The Thirty-Third AAAI Conference on Artificial Intelligence (AAAI-19) cosigned a statement with CRA, and the Thirty-First Conference on Innovative Applications of Artificial expressing concern about the proposed Intelligence (IAAI-19), will be held in Honolulu, Hawaii, USA, January tax bill and its ramifications for graduate 27 - February 1, 2019. The technical conference will continue its student stipends. Other organizational 3.5-day schedule, preceded by the workshop and tutorial programs.

Active Reinforcement Learning (ARL) is a twist on RL where the agent observes reward information only if it pays a cost. This subtle change makes exploration substantially more challenging. Powerful principles in RL like optimism, Thompson sampling, and random exploration do not help with ARL. We relate ARL in tabular environments to Bayes-Adaptive MDPs. We provide an ARL algorithm using Monte-Carlo Tree Search that is asymptotically Bayes optimal. Experimentally, this algorithm is near-optimal on small Bandit problems and MDPs. On larger MDPs it outperforms a Q-learner augmented with specialised heuristics for ARL. By analysing exploration behaviour in detail, we uncover obstacles to scaling up simulation-based algorithms for ARL.

Deep reinforcement learning has been successfully applied to several visual-input tasks using model-free methods. In this paper, we propose a model-based approach that combines learning a DNN-based transition model with Monte Carlo tree search to solve a block-placing task in Minecraft. Our learned transition model predicts the next frame and the rewards one step ahead given the last four frames of the agent's first-person-view image and the current action. Then a Monte Carlo tree search algorithm uses this model to plan the best sequence of actions for the agent to perform. On the proposed task in Minecraft, our model-based approach reaches the performance comparable to the Deep Q-Network's, but learns faster and, thus, is more training sample efficient. Keywords: Acknowledgements Reinforcement Learning, Model-Based Reinforcement Learning, Deep Learning, Model Learning, Monte Carlo Tree Search I would like to express my sincere gratitude to my supervisor Dr. Stefan Uhlich for his continuous support, patience, and immense knowledge that helped me a lot during this study. My thanks and appreciation also go to my colleague Anna Konobelkina for insightful comments on the paper as well as to Sony Europe Limited for providing the resources for this project.

Learning is an important aspect of human intelligence. People learn from various aspects of their experience over time. We present an episodic infrastructure for learning in the context of a cognitive architecture, \icarus/. After a review of this architecture, we formally define the architectural extensions for episodic capabilities. We then demonstrate the extended system's capability to learn planning operators using the episodic traces from two Minecraft-like scenarios.