Modeling competitive situations

This chapter studies how competitive situations are conventionally modeled in noncooperative game theory. It uses two sorts or forms of models: the so-called extensive form game and the normal or strategic form game. An extensive form representation of a noncooperative game is composed of the following list of items: a list of players; a game tree; an assignment of decision nodes to players or to nature; lists of actions available at each decision node and a correspondence between immediate successors of each decision node and available actions; information sets; an assignment of payoffs for each player to terminal nodes; and probability assessments over the initial nodes and over the actions at any node that is assigned to nature. There is no single way to proceed in general from a normal form game to a corresponding extensive form game. In one obvious extensive form, the players all choose complete strategies simultaneously, but often other extensive forms could be constructed from a given normal form.

Next decision node (NDN) planning: an ambidextrous planning model

Purpose The purpose of this paper is to address calls from within the existing literature for a planning model that can provide both control and flexibility in dynamic environments. Design/methodology/approach This research adopts a two-phase grounded theory methodology. Phase I tests a hypothesis through a structured online survey. Phase II tests a new planning model through a focus group and an online survey. Findings This research finds that client-side project managers misapply rational planning tools when managing construction projects. It also finds that the planning artefacts created in the initial stages of a project are used as a communication tool, rather than a controlling or monitoring tool. These findings provide an opportunity for new planning models, such as the (next decision node) NDN model, to be adopted. These planning tools can create new levels of transparency and accountability within the planning process – as well as provide a model which is more closely aligned with the practitioners “lived experience”. Research limitations/implications The NDN model provides a foundation for further research into the micro-mechanisms of organisational ambidexterity in projects and refinement of the NDN model. Practical implications This research presents a new planning model which practitioners can adopt when delivering construction projects. Originality/value This paper presents a new, ambidextrous planning model grounded on the “lived experience” of client-side project managers.

DeltaDou: Expert-level Doudizhu AI through Self-play

Artificial Intelligence has seen several breakthroughs in two-player perfect information game. Nevertheless, Doudizhu, a three-player imperfect information game, is still quite challenging. In this paper, we present a Doudizhu AI by applying deep reinforcement learning from games of self-play. The algorithm combines an asymmetric MCTS on nodes of information set of each player, a policy-value network that approximates the policy and value on each decision node, and inference on unobserved hands of other players by given policy. Our results show that self-play can significantly improve the performance of our agent in this multi-agent imperfect information game. Even starting with a weak AI, our agent can achieve human expert level after days of self-play and training.

The Alternative Net Model with the Fuzzy Decision Node for the Construction Projects Planning

The new approach to the construction project planning is presented in the article. The classical net model is enriched by the fuzzy decision node. The decision node allows for alternate choices dependent on appearing circumstances. The alternative net model with fuzzy decision node is an acyclic multi-graph, where some, chosen nodes (events) have multiple connections. These connections represent alternative methods of the certain work execution. Every work (activity) (i, j) in the net model with alternative methods of work execution, despite the basic information comprising the execution time, the cost, the number of necessary workers, should comprise additional information e.g. about a complexity of works, a real feasibility. The alternative ways of a given work execution are evaluated in the decision node based on the fuzzy decision model. Each method is evaluated by assigning it the preference level in a form of the value of the membership function - „equal or higher” μ≥ . The most preferable way of the work execution in a given circumstances, will have the highest value of preference level. When the choice is done the net model is solved in the traditional way. Therefore, the paper concentrates on the process of choosing the method of work execution in the fuzzy decision node. The example calculations accompanying the process of decision taking are presented too. The model requires the use of linguistic variables, a fuzzy numbers, as well as fuzzy preference relations together with some calculations applied the probability theory.

Route Choice in Subway Station during Morning Peak Hours: A Case of Guangzhou Subway

This paper is aimed at crowding phenomenon in the subway. As passengers are inclined to choose the route with minimum disutility, we put forward a route choice model which is constructed to achieve minimum objective function of feasibility for the optimal solution. Meanwhile we set passenger volume threshold values according to capacity of facilities. In the case of actual capacity exceeding the threshold, the decision node of constrained route will be selected; computing procedure about searching decision points will be presented. Then we should set rational restrictions at the decision node of the minimum utility function route to prevent too many passengers’ access to platform. Through certification, this series of methods can effectively ensure the safety of the station efficient operation.

Inducting Fuzzy Decision Tree Based on Discrete Attributes through Uncertainty Reduction

In this paper, we give a fuzzy decision tree (simply FDT) induction algorithm, named FDTAmbig, to handle the classification with discrete attributes through the uncertainty reduction. In FDTAmbig, the uncertainty is measured with classification ambiguity. FDTAmbig selects the attribute which will cause the further reduction of uncertainty as the expanded attribute for each decision node. The experimental result shows that FDTAmbig has the better generalization capability in comparison with the FDT induced with classification entropy (FDTEntr).