The presented diploma thesis examines the possibilities of developing algorithmic thinking in primary school pupils. Algorithmic thinking is a necessary tool for effective analysis of the problem and the creation of a procedure for its subsequent repeatable solution. The main research topic of this work is effective and efficient ways of developing algorithmic thinking, especially how to formulate a problem and how to develop algorithmic thinking with the greatest effect in the educational process. The work summarizes various ways of developing algorithmic thinking and the approach to its teaching. The concepts of computer thinking, algorithmic thinking, algorithm and algorithmization are defined and analyzed. Furthermore, contemporary means of developing algorithmic thinking described in the literature are specified. They are then critically evaluated and at the end of the theoretical part the most suitable set of means is selected and the way of their use is proposed. Action research was carried out in the circle of informatics at the primary school in Prague. The verification of the effectiveness of the selected instruments took place during fifteen sixty-minute lessons. The focus of this research lies in increasing the quality of pedagogical practice of the teacher, in the development of his didactic...

No description available.

This thesis is concerned with policy space abstractions that concisely encode alternative ways of making decisions; dealing with discovery, learning, adaptation and use of these abstractions. This work is motivated by the problem faced by autonomous agents that operate within a domain for long periods of time, hence having to learn to solve many different task instances that share some structural attributes. An example of such a domain is an autonomous robot in a dynamic domestic environment. Such environments raise the need for transfer of knowledge, so as to eliminate the need for long learning trials after deployment. Typically, these tasks would be modelled as sequential decision making problems, including path optimisation for navigation tasks, or Markov Decision Process models for more general tasks. Learning within such models often takes the form of online learning or reinforcement learning. However, handling issues such as knowledge transfer and multiple task instances requires notions of structure and hierarchy, and that raises several questions that form the topic of this thesis – (a) can an agent acquire such hierarchies in policies in an online, incremental manner, (b) can we devise mathematically rigorous ways to abstract policies based on qualitative attributes, (c) when it is inconvenient to employ prolonged trial and error learning, can we devise alternate algorithmic methods for decision making in a lifelong setting? The first contribution of this thesis is an algorithmic method for incrementally acquiring hierarchical policies. Working with the framework of options - temporally extended actions - in reinforcement learning, we present a method for discovering persistent subtasks that define useful options for a particular domain. Our algorithm builds on a probabilistic mixture model in state space to define a generalised and persistent form of ‘bottlenecks’, and suggests suitable policy fragments to make options. In order to continuously update this hierarchy, we devise an incremental process which runs in the background and takes care of proposing and forgetting options. We evaluate this framework in simulated worlds, including the RoboCup 2D simulation league domain. The second contribution of this thesis is in defining abstractions in terms of equivalence classes of trajectories. Utilising recently developed techniques from computational topology, in particular the concept of persistent homology, we show that a library of feasible trajectories could be retracted to representative paths that may be sufficient for reasoning about plans at the abstract level. We present a complete framework, starting from a novel construction of a simplicial complex that describes higher-order connectivity properties of a spatial domain, to methods for computing the homology of this complex at varying resolutions. The resulting abstractions are motion primitives that may be used as topological options, contributing a novel criterion for option discovery. This is validated by experiments in simulated 2D robot navigation, and in manipulation using a physical robot platform. Finally, we develop techniques for solving a family of related, but different, problem instances through policy reuse of a finite policy library acquired over the agent’s lifetime. This represents an alternative approach when traditional methods such as hierarchical reinforcement learning are not computationally feasible. We abstract the policy space using a non-parametric model of performance of policies in multiple task instances, so that decision making is posed as a Bayesian choice regarding what to reuse. This is one approach to transfer learning that is motivated by the needs of practical long-lived systems. We show the merits of such Bayesian policy reuse in simulated real-time interactive systems, including online personalisation and surveillance.

We investigate the properties of a new Multi-Agent Systems (MAS) for computer-aided composition called IPCS (pronounced “ipp-siss”) the Intermediate Performance Composition System which generates expressive performance as part of its compositional process, and produces emergent melodic structures by a novel multi-agent process. IPCS consists of a small-medium size (2 to 16) collection of agents in which each agent can perform monophonic tunes and learn monophonic tunes from other agents. Each agent has an affective state (an “artificial emotional state”) which affects how it performs the music to other agents; e.g. a “happy” agent will perform “happier” music. The agent performance not only involves compositional changes to the music, but also adds smaller changes based on expressive music performance algorithms for humanization. Every agent is initialized with a tune containing the same single note, and over the interaction period longer tunes are built through agent interaction. Agents will only learn tunes performed to them by other agents if the affective content of the tune is similar to their current affective state; learned tunes are concatenated to the end of their current tune. Each agent in the society learns its own growing tune during the interaction process. Agents develop “opinions” of other agents that perform to them, depending on how much the performing agent can help their tunes grow. These opinions affect who they interact with in the future. IPCS is not a mapping from multi-agent interaction onto musical features, but actually utilizes music for the agents to communicate emotions. In spite of the lack of explicit melodic intelligence in IPCS, the system is shown to generate non-trivial melody pitch sequences as a result of emotional communication between agents. The melodies also have a hierarchical structure based on the emergent social structure of the multi-agent system and the hierarchical structure is a result of the emerging agent social interaction structure. The interactive humanizations produce micro-timing and loudness deviations in the melody which are shown to express its hierarchical generative structure without the need for structural analysis software frequently used in computer music humanization.

We used a divide-and-conquer algorithm to recursively solve the two-dimensional problem of protein folding of an HP sequence with the maximum number of H-H contacts. We derived both lower and upper bounds for the algorithmic complexity by using the newly introduced concept of multi-directional width-bounded geometric separator. We proved that for a grid graph G with n grid points P, there exists a balanced separator A subseteq P$ such that A has less than or equal to 1.02074 sqrt{n} points, and G-A has two disconnected subgraphs with less than or equal to {2over 3}n nodes on each subgraph. We also derive a 0.7555sqrt {n} lower bound for our balanced separator. Based on our multidirectional width-bounded geometric separator, we found that there is an O(n^{5.563sqrt{n}}) time algorithm for the 2D protein folding problem in the HP model. We also extended the upper bound results to rectangular and triangular lattices.

Building automated trading systems has long been one of the most cutting-edge and exciting fields in the financial industry. In this research project, we built a trading system based on machine learning methods. We used the Recurrent Reinforcement Learning (RRL) algorithm as our fundamental algorithm, and by introducing Genetic Algorithms (GA) in the optimization procedure, we tackled the problems of picking good initial values of parameters and dynamically updating the learning speed in the original RRL algorithm. We call this optimization algorithm the Evolutionary Recurrent Reinforcement Learning algorithm (ERRL), or the GA-RRL algorithm. ERRL allows us to find many local optimal solutions easier and faster than the original RRL algorithm. Finally, we implemented the GA-RRL system on EUR/USD at a 5-minute level, and the backtest performance showed that our GA-RRL system has potentially promising profitability. In future research we plan to introduce some risk control mechanism, implement the system on different markets and assets, and perform backtest at higher frequency level.

Thesis (M.M. (Finance & Investment))--University of the Witwatersrand, Faculty of Commerce, Law and Management, Graduate School of Business Administration, 2013. / The evidence put forward by Zhang (2010) indicates that algorithmic trading can potentially generate the momentum effect evident in empirical market research. In addition, upon analysis of the literature, it is apparent that algorithmic traders possess a comparative informational advantage relative to regular traders. Finally, the theoretical model proposed by Wang (1993), indicates that the informational differences between traders fundamentally influences the nature of asset prices, even generating serial return correlations. Thus, applied to the study, the theory holds that algorithmic trading would have a significant effect on security return dynamics, possibly even engendering the momentum effect. This paper tests such implications by proposing a theory to explain the momentum effect based on the hypothesis that algorithmic traders possess Innovative Information about a firm’s future performance. From this perspective, Innovative Information can be defined as the information derived from the ability to accumulate, differentiate, estimate, analyze and utilize colossal quantities of data by means of adept techniques, sophisticated platforms, capabilities and processing power. Accordingly, an algorithmic trader’s access to various complex computational techniques, infrastructure and processing power, together with the constraints to human information processing, allow them to make judgments that are superior to the judgments of other traders. This particular aspect of algorithmic trading remains, to the best of my knowledge, unexplored as an avenue or mechanism, through which algorithmic trading could possibly affect the momentum effect and thus market efficiency. Interestingly, by incorporating this information variable into a simplified representative agent model, we are able to produce return patterns consistent with the momentum effect in its entirety. The general thrust of our results, therefore, is that algorithmic trading can hypothetically generate the return anomaly known as the momentum effect. Our results give credence to the assumption that algorithmic trading is having a detrimental effect on stock market efficiency.

Dans cette thèse, nous présentons une approche algorithmique permettant la génération de guides de construction de cages moléculaires organiques. Il s'agit d'architectures semi-moléculaires possédant un espace interne défini capable de piéger une molécule cible appelée substrat. De nombreuses œuvres proposent de générer des cages organiques moléculaires obtenues à partir de structures symétriques, qui ont une bonne complexité, mais elles ne sont pas spécifiques car elles ne prennent pas en compte des cibles précises. L'approche proposée permet de générer des guides de construction de cages moléculaires organiques spécifiques à un substrat donné. Afin de garantir la spécificité de la cage moléculaire pour le substrat cible, une structure intermédiaire, qui est une expansion de l'enveloppe du substrat cible, est utilisée. Cette structure définie la forme de l'espace dans lequel est piégé le substrat. Des petits ensembles d'atomes, appelés motifs moléculaires liants, sont ensuite intégrés à cette structure intermédiaire. Ces motifs moléculaires sont les ensembles d'atomes nécessaires aux cages moléculaires pour leur permettre d’interagir avec le substrat afin de le capturer. / In this thesis, we present an algorithmic approach allowing the generation of construction guides of organic molecular cages. These semi-molecular architectures have a defined internal space capable of trapping a target molecule called substrate. Many works propose to generate molecular organic cages obtained from symmetrical structures, which have a good complexity, but they are not specific because they do not take into account precise targets. The proposed approach makes it possible to generate guides for the construction of organic molecular cages specific to a given substrate. In order to ensure the specificity of the molecular cage for the target substrate, an intermediate structure, which is an expansion of the envelope of the target substrate, is used. This structure defines the shape of the space in which the substrate is trapped. Small sets of atoms, called molecular binding patterns, are then integrated into this intermediate structure. These molecular patterns are the sets of atoms needed by molecular cages to allow them to interact with the substrate to capture it.

Esta dissertação tem por objetivo apresentar uma abordagem baseada em técnicas de inteligência artificial para automatizar a etapa de particionamento de modelos em simulação distribuída. Essa abordagem utiliza os conceitos da computação evolutiva para o desenvolvimento de um algoritmo genético capaz de otimizar o processo de particionamento e auxiliar a tomada de decisões na tarefa de obtenção dos processos lógicos. Objetiva-se com sua aplicação minimizar o tempo de execução da simulação distribuída, evitando que o pior tempo de execução seja utilizado. Para alcançar esse objetivo, o particionamento apresentado como solução é caracterizado pelo balanceamento de carga e pela baixa latência de comunicação entre processos. Isso é possível porque o algoritmo genético utiliza informações contidas no modelo e na arquitetura de onde a simulação será executada. Esses padrões são utilizados para obter informações sobre a comunicação entre processos, a carga de processamento por centro de serviço e a capacidade de processamento das máquinas / This dissertation presents an approach based on intelligence artificial technics to automatize the model partitioning stage in distributed simulation. This approach makes uses evolutive computing concepts to developed a genetic algorithmic that can optimize the partitioning process and help to take decisions in the task to get the logical process. The propose of this algorithm is reduce to execution time the distributed simulation and to avoid the use of the worst execution time. To reach this target, the partitioning obtained has characteristics such as load balance and the low-communication interprocess. This is possible because the genetic algorithmic uses as input information from the model and the architect where the simulation with be executed. These inputs are used to get information about the interprocess communication, processing load per service center and processing capacity in the machines

This thesis is about algorithmic creation in the arts - where an artist, designer or composer uses a formal generative process to assist in crafting forms and patterns - and approaches to finding effective input parameter values to these generative processes for aesthetic ends. Framed in three practical studies, approaches to navigating the aesthetic possibilities of generative processes in sound and visuals are presented, and strategies for eliciting the preferences of the consumers of the generated output are explored. The first study presents a musical interface that enables navigation of the possibilities of a stochastic generative process with respect to measures of subjective predictability. Through a mobile phone version of the application, aesthetic preferences are crowd-sourced. The second study presents an eye-tracking based framework for the exploration of the possibilities afforded by generative designs; the interaction between the viewers' gaze patterns and the system engendering a fluid navigation of the state-space of the visual forms. The third study presents a crowd-sourced interactive evolutionary system, where populations of abstract colour images are shaped by thousands of preference selections from users worldwide For each study, the results of analyses eliciting the attributes of the generated outputs - and their associated parameter values - that are most preferred by the consumers/users of these systems are presented. Placed in a historical and theoretical context, a refined perspective on the complex interrelationships between generative processes, input parameters and perceived aesthetic value is presented. Contributions to knowledge include identified trends in objective aesthetic preferences in colour combinations and their arrangements, theoretical insights relating perceptual mechanisms to generative system design and analysis, strategies for effectively leveraging evolutionary computation in an empirical aesthetic context, and a novel eye-tracking based framework for the exploration of visual generative designs.