Key to reinforcement learning in multi-agent systems is the ability to exploit the fact that agents only directly influence only a small subset of the other agents. Such loose couplings are often modelled using a graphical model: a coordination graph. Finding an (approximately) optimal joint action for a given coordination graph is therefore a central subroutine in cooperative multi-agent reinforcement learning (MARL). Much research in MARL focuses on how to gradually update the parameters of the coordination graph, whilst leaving the solving of the coordination graph up to a known typically exact and generic subroutine. However, exact methods { e.g., Variable Elimination { do not scale well, and generic methods do not exploit the MARL setting of gradually updating a coordination graph and recomputing the joint action to select. In this paper, we examine what happens if we use a heuristic method, i.e., local search, to select joint actions in MARL, and whether we can use outcome of this local search from a previous time-step to speed up and improve local search. We show empirically that by using local search, we can scale up to many agents and complex coordination graphs, and that by reusing joint actions from the previous time-step to initialise local search, we can both improve the quality of the joint actions found and the speed with which these joint actions are found.
LINK
Artificially intelligent agents increasingly collaborate with humans in human-agent teams. Timely proactive sharing of relevant information within the team contributes to the overall team performance. This paper presents a machine learning approach to proactive communication in AI-agents using contextual factors. Proactive communication was learned in two consecutive experimental steps: (a) multi-agent team simulations to learn effective communicative behaviors, and (b) human-agent team experiments to refine communication suitable for a human team member. Results consist of proactive communication policies for communicating both beliefs and goals within human-agent teams. Agents learned to use minimal communication to improve team performance in simulation, while they learned more specific socially desirable behaviors in the human-agent team experiment
DOCUMENT
Research into the relationship between innovative physical learning environments (PLEs) and innovative psychosocial learning environments (PSLEs) indicates that it must be understood as a network of relationships between multiple psychosocial and physical aspects. Actors shape this network by attaching meanings to these aspects and their relationships in a continuous process of gaining and exchanging experiences. This study used a psychosocial-physical, relational approach for exploring teachers’ and students’ experiences with six innovative PLEs in a higher educational institute, with the application of a psychosocial-physical relationship (PPR) framework. This framework, which brings together the multitude of PLE and PSLE aspects, was used to map and analyse teachers’ and students’ experiences that were gathered in focus group interviews. The PPR framework proved useful in analysing the results and comparing them with previous research. Previously-identified relationships were confirmed, clarified, and nuanced. The results underline the importance of the attunement of system aspects to pedagogical and spatial changes, and of a psychosocial-physical relational approach in designing and implementing new learning environments, including the involvement of actors in the discourse within and between the different system levels. Interventions can be less invasive, resistance to processes could be reduced, and innovative PLEs could be used more effectively.
MULTIFILE
We present a novel architecture for an AI system that allows a priori knowledge to combine with deep learning. In traditional neural networks, all available data is pooled at the input layer. Our alternative neural network is constructed so that partial representations (invariants) are learned in the intermediate layers, which can then be combined with a priori knowledge or with other predictive analyses of the same data. This leads to smaller training datasets due to more efficient learning. In addition, because this architecture allows inclusion of a priori knowledge and interpretable predictive models, the interpretability of the entire system increases while the data can still be used in a black box neural network. Our system makes use of networks of neurons rather than single neurons to enable the representation of approximations (invariants) of the output.
LINK
Dit essay geeft een systeemvisie op het ontwikkelen van embedded software voor slimme systemen: (mobiele) robots en sensornetwerken.
DOCUMENT
Peer-to-peer (P2P) energy trading has been recognized as an important technology to increase the local self-consumption of photovoltaics in the local energy system. Different auction mechanisms and bidding strategies haven been investigated in previous studies. However, there has been no comparatively analysis on how different market structures influence the local energy system’s overall performance. This paper presents and compares two market structures, namely a centralized market and a decentralized market. Two pricing mechanisms in the centralized market and two bidding strategies in the decentralized market are developed. The results show that the centralized market leads to higher overall system self-consumption and profits. In the decentralized market, some electricity is directly sold to the grid due to unmatchable bids and asks. Bidding strategies based on the learning algorithm can achieve better performance compared to the random method.
DOCUMENT
In most European countries entrepreneurship is one of the top priorities on the national agenda, to stimulate individual and organizational innovativeness and (regional) economic growth. As a consequence, embedding entrepreneurship in education to achieve this goal has gained importance and momentum especially at universities of applied sciences. Two questions need answering when trying to embed entrepreneurship in a curriculum. First of all: cán entrepreneurship be taught and second: hów should entrepreneurship be taught. In this paper we focus on an educational programme based on a learner-cantered, constructivist approach, which is offered in a multidisciplinary, inspiring and entrepreneurial setting. It is competency-based and is tailor-made to individual student demand and goes beyond the classic business school approach based on instruction. The programme caters for students from at least 40 different departments of the university. The starting point in this programme is the assumption that entrepreneurship can indeed be taught but that the pedagogical climate and approach is crucial and should contribute towards the development of entrepreneurial competencies and skills. In this paper issues such as the dynamics of learning are dealt with as well as some a discussion on learning paradigms. We elaborate on the programme developed at The Hague University of Applied Sciences, The Hague in The Netherlands. So far, over 250 students have participated in the programme and since September 2007 longitudinal research has taken place to establish the effects of the programme and the pedagogical approach on the development of entrepreneurialism. We then describe the research design and draw preliminary conclusions about the relation between pedagogical climate and entrepreneurial behaviour, competencies and entrepreneurial behaviour and finally the relation between entrepreneurial behaviour and the choice to become an independent entrepreneur. Our findings show that such competencies as self-discipline and vulnerability are positive influencers of entrepreneurial ambition. We also found negative influencers of entrepreneurial ambition in depression and inadequacy, yet interestingly also in sincerity. The role of the business partners involved in the programme is discussed and an account is given of the experiences of a population of students over a period of three years on the basis of a number of issues: what works, what doesn't work and what needs to be improved. Interesting drivers for entrepreneurial behaviour are distilled from our research, on the basis of which recommendations are given on how to best implement these drivers into an educational programme. The paper finalizes with a concluding note in which some of the drawbacks of a learner-centred approach as opposed to an instruction-based approach are discussed and suggestions for future research are made.
DOCUMENT
A case study and method development research of online simulation gaming to enhance youth care knowlegde exchange. Youth care professionals affirm that the application used has enough relevance as an additional tool for knowledge construction about complex cases. They state that the usability of the application is suitable, however some remarks are given to adapt the virtual environment to the special needs of youth care knowledge exchange. The method of online simulation gaming appears to be useful to improve network competences and to explore the hidden professional capacities of the participant as to the construction of situational cognition, discourse participation and the accountability of intervention choices.
DOCUMENT
Robot tutors provide new opportunities for education. However, they also introduce moral challenges. This study reports a systematic literature re-view (N = 256) aimed at identifying the moral considerations related to ro-bots in education. While our findings suggest that robot tutors hold great potential for improving education, there are multiple values of both (special needs) children and teachers that are impacted (positively and negatively) by its introduction. Positive values related to robot tutors are: psychological welfare and happiness, efficiency, freedom from bias and usability. However, there are also concerns that robot tutors may negatively impact these same values. Other concerns relate to the values of friendship and attachment, human contact, deception and trust, privacy, security, safety and accountability. All these values relate to children and teachers. The moral values of other stakeholder groups, such as parents, are overlooked in the existing literature. The results suggest that, while there is a potential for ap-plying robot tutors in a morally justified way, there are imported stake-holder groups that need to be consulted to also take their moral values into consideration by implementing tutor robots in an educational setting. (from Narcis.nl)
MULTIFILE
Relatief kleine, gespecialiseerde bedrijven in de maakindustrie hebben behoefte aan flexibele assemblageprocessen en productielogistiek. Digitalisering biedt veel mogelijkheden om productieprocessen efficiënter en duurzamer te maken, innovatieve producten te fabriceren en over te schakelen op andere businessmodellen. Dit moet dan wel werken voor kleine series en enkelstuks. ‘Kunnen wij het maken?’ verwijst naar onderliggende vragen over: ‘Hoe beheersen we risico’s in complexe maakprocessen?’, ‘Hoe werken we samen in de keten?’ en ‘Wat moeten huidige en toekomstige engineers weten over ‘Industry 4.0’ en circulaire maakindustrie?’. Bijgaand essay, in verkorte vorm uitgesproken als Intreerede, legt uit hoe de onderzoekers van Smart Sustainable Manufacturing aan de slag gaan om een antwoord te vinden op deze vragen, door middel van cocreatie met de beroepspraktijk en het onderwijs in het Re/manufacturing lab.
DOCUMENT
