After rigorous criticism on trait approach and with the emergence of behavioral approach in entrepreneurship during 1980s, the researchers started to introduce learning and cognitive theories in entrepreneurship to describe and explain the dynamic nature of entrepreneurship. Many researchers have described venture creation as a core and the single most important element of entrepreneurship. This thesis will discuss and present the role of entrepreneurial learning and heuristics in venture creation. Hence, the purpose of this research thesis is to study and analyze the role of entrepreneurial learning and heuristics in venture creation. To fulfill the purpose of this thesis, we followed qualitative research and conducted semi structured interviews with open ended questionnaires to collect empirical data. For this study, we have included only four interviews which were conducted on four different businesses based in Jönköping, Sweden, following convenience sampling. In the analysis, we used data analysis model of Walker, Cooke and McAllister (2008) and inductively generated three propositions, depicting the role and importance of entrepreneurial learning and heuristics in venture creation. Individuals adopt entrepreneurship in their careers with necessary skills, abilities, and knowledge, which are learned or gained through experiential learning and/or vicarious learning (i.e., learning by observing or modeling the actions of others). Learning by doing is considered the most important factor by entrepreneurs which helped them to overcome different business start up hurdles, to make various entrepreneurial decisions and to perform many entrepreneurial activities during venture creation. Similarly, individuals within their own situation use, learning by observing or modeling other people’s behaviour, actions and consequences of the actions. Entrepreneurs use learning by modeling the behaviour and actions of others as benchmarking strategy during venture creation. Entrepreneurs believe that without any learning they will not be able to start their own businesses. Heuristics as decisions making mechanism, particularly during venture creation, is used by entrepreneurs as simplifying strategy when sufficient information related to a specific market, certain industry and products are scarce. Additionally, entrepreneurs are passionate to grab profitable business opportunity, and due to time pressure and brief window of opportunity, they can’t go for gathering each and every information of the potential business or product. Hence, heuristics as decisions making mechanism is considered the best suitable approach to make many entrepreneurial decisions during venture creation.

Machine learning in natural language has been a widely pursued area of research. However, few learning techniques model themselves after human learning, despite the nature of the task being closely connected to human cognition. In particular, the idea of learning language in stages is a common approach for human learning, as can be seen in practice in the education system and in research on language acquisition. However, staged learning for natural language is an area largely overlooked by machine learning researchers. This thesis proposes a developmental learning heuristic for natural language models, to evaluate its performance on natural language tasks. The heuristic simulates human learning stages by training on child, teenage and adult text, provided by the British National Corpus. The three staged learning techniques that are proposed take advantage of these stages to create a single developed Hidden Markov Model. This model is then applied to the task of part-of-speech tagging to observe the effects of development on language learning.

We consider an abstract representation of some environment in which an agent is located. Given a goal sequence, we ask what strategy said agent - utilizing readily available algorithmic tools - should incorporate to successfully find a valid traversal route such that it is optimal in accordance with a predefined error-margin. We present four scenarios that each incorporate aspects common to general navigation to further illustrate some of the difficult problems needed to be solved in any general navigation strategy. Two reinforcement learning and four graph path planning algorithms are studied and applied on said predefined scenarios. Through the introduction of a long-term strategy model we allow comparative study of the result of the applications, and note a distinct difference in performance. Further, we discuss the lack of a probabilistic algorithmic approach and why it should be an option in any general strategy as it allows verifiably "good" estimated solutions, useful when the problem at hand is NP-hard. Several meta-level concepts are introduced and discussed to further illustrate the difficulty in producing an optimal strategy with an explicit long-term horizon. We argue for a non-deterministic approach, looking at the apparent gain of epsilon-randomness when incorporated by a reinforcement learning agent. Several problems that may arise with non-determinism are discussed, based on the notion that such an agents' performance can be viewed as a markov chain; possibly resulting in suboptimal paths concerning norm.

Syftet med denna studie är att undersöka hur elever samtalar när de möter en uppgift som syftar till att öva deras förmåga att använda ett naturvetenskapligt arbetssätt. Undersökningen genomfördes genom att fyra grupper om två elever fick en uppgift där de formulerade frågeställningar kring en isballong (en frusen vattenballong). Samtalen spelades in på band. Inspelningarna transkriberades och analyserades med hjälp av en praktisk epistemologisk analys utifrån tre olika kategorier: samtal inom diskursen, samtal om diskursen och samtal utanför diskursen. Resultatet av undersökningen visar att de flesta eleverna har, trots att de har ringa eller ingen erfarenhet av att arbeta med natruvetenskapligt arbetssätt, ganska lätt att ta till sig uppgiften. Resultatet visar även att det inte är samtal som ligger utanför ramen för uppgiften som utgör det största hindret för eleverna att arbeta med uppgiften, utan i stället att det är samtal som rör uppgiftens utformning.

This Master’s Thesis deals with a special problem that may be of importance when planning a frequency hopping mobile communication network. In normal cases the Frequency Assignment Problem is solved, in order to plan the use of frequencies in a network. The special case discussed in this thesis occurs when the network operator requires that the frequencies must be arranged into groups. In this case the Frequency Assignment Problem must be solved with respect to the groups, i.e. a Group assignment Problem. The thesis constitutes the final part of the Master of Science in Communication and Transport Systems Engineering education, at Linköping University, Campus Norrköping. The Group Arrangement Problem was presented by ComOpt, a company that has specialized in solving the Frequency Assignment Problem for network operators. This thesis does not deal with solutions for the Frequency Assignment Problem, with respect to the groups. The main issue in the thesis is to construct a computer based algorithm that solves the Group Arrangement Problem, i.e. creating the groups. The goal is to construct an algorithm that creates groups which imply a better solution for the Frequency Assignment Problem than manually created groups. Two algorithms are presented and tested on two cases. Their respective results for both cases are compared with the results from a manual grouping. The two computer based algorithms creates better groups than the manual grouping strategy, according to an artificial quality measure. As of spring 2003 a variant of one of the presented algorithms was implemented in ComOpt’s product for solving the Frequency Assignment Problem.

Machine learning in natural language has been a widely pursued area of research. However, few learning techniques model themselves after human learning, despite the nature of the task being closely connected to human cognition. In particular, the idea of learning language in stages is a common approach for human learning, as can be seen in practice in the education system and in research on language acquisition. However, staged learning for natural language is an area largely overlooked by machine learning researchers. This thesis proposes a developmental learning heuristic for natural language models, to evaluate its performance on natural language tasks. The heuristic simulates human learning stages by training on child, teenage and adult text, provided by the British National Corpus. The three staged learning techniques that are proposed take advantage of these stages to create a single developed Hidden Markov Model. This model is then applied to the task of part-of-speech tagging to observe the effects of development on language learning.

The fair-value hierarchy used in financial accounting has been criticized because of its complexity being the reason for several accounting issues. This study examines the underlying factors affecting decision makers in the process of fair-value accounting of financial instruments within the fair-value hierarchy. Research has been conducted through in-depth interviews with representatives of five Swedish banks. The findings have been analysed with a frame of reference built on prior judgment and decision making research. The results of the study show that the extent of the affecting factors vary between different banks depending on the banks' individual prerequisites.

A fundamental issue in warehouse operations is the storage location of the products it contains. Placing products intelligently within the system can allow for great reductions in order pick costs. This is essential because order picking is a major cost of warehouse operations. For example, a study by Drury conducted in the UK found that 63% of warehouse operating costs are due to order picking. When orders contain a single item, the COI rule of Heskett is an optimal storage policy. This is not true when orders contain multiple line items because no information is used about what products are ordered together. In this situation, products that are frequently ordered together should be stored together. This is the basis of the correlated storage assignment problem. Several previous researchers have considered how to form such clusters of products with an ultimate objective of minimizing travel time. In this dissertation, we focus on the alternate objective of minimizing multi-zone orders. We present a mathematical model and discuss properties of the problem. A Lagrangian relaxation solution approach is discussed. In addition, we both develop and adapt several heuristics from the literature to give upper bounds for the model. A cyclic exchange improvement method is also developed. This exponential size neighborhood can be efficiently searched in polynomial time. Even for poor initial solutions, this method finds solutions which outperform the best approaches from the literature. Different product sizes, stock splitting, and rewarehousing are problem features that our model can handle. The cyclic exchange algorithm is also modified to allow these operating modes. In particular, stock splitting is a difficult issue which most previous research in correlated storage ignores. All of our algorithms are implemented and tested on data from a functioning warehouse. For all data sets, the cyclic exchange algorithm outperforms COI, the standard industry approach, by an average of 15%.

This dissertation focuses on improvements in vehicle routing that can be gained by allowing multiple vehicles to service a common load. We explore how costs can be reduced through the elimination of the constraint that a load must be serviced by only one vehicle. Specifically, we look at the problem of routing vehicles to service loads that have distinct origins and destinations, with no constraint on the amount of a load that a vehicle may service. We call this the Pickup and Delivery Problem with Split Loads (PDPSL). We model this problem as a dynamic program and introduce structural results that can help practitioners implement the use of split loads, including the definition of an upper bound on the benefit of split loads. This bound indicates that the routing cost can be reduced by at most one half when split loads are allowed. Furthermore, the most benefit occurs when load sizes are just above one half of vehicle capacity. We develop a heuristic for the solution of large scale problems, and apply this heuristic to randomly generated data sets. Various load sizes are tested, with the experimental results supporting the finding that most benefit with split loads occurs for load sizes just above one half vehicle capacity. Also, the average benefit of split loads is found to range from 6 to 7% for most data sets. The heuristic was also tested on a real world example from the trucking industry. These tests reveal the benefit of both using split loads and allowing fleet sharing. The benefit for split loads is not as significant as with the random data, and the various business rules added for this case are tested to find those that have the most impact. It is found that an additional cost for every stop the vehicle makes strictly limits the potential for benefit from split loads. Finally, we present a simplified version of the PDPSL in which all origins are visited prior to any destination on a route, generalizing structural results from the Split Delivery Vehicle Routing Problem for this problem.

People are increasingly concerned about online privacy and how computers collect, process, share, and store their personal information. Such concerns are understandable given the growing number of privacy invasions and the pervasiveness of information capture and sharing between IT systems. This situation has led to an increasingly regulated environment, limiting what systems may do, and what safeguards they must offer users. Privacy is an especially important concern in the fields of computer supported collaborative work (CSCW), Ubiquitous Computing, and e-commerce, where the nature of the applications often requires some information collection and sharing. In order to minimize risks to users it is essential to identify privacy problems early in the design process. Several methods and frameworks for accomplishing this have been proposed in the last decades. These frameworks, though based on hard-earned experience and great insight, have not seen widespread adoption despite the high level of interest in this topic. Part of the reason for this is likely the lack of evaluation and study of these frameworks. In our research we examine the key design and analysis frameworks and their elements, and compare these to the kinds of problems users face and are concerned with in terms of privacy. Based on this analysis of the relative strengths and weaknesses of existing design frameworks we derive a new design framework; STRAP (STRuctured Analysis of Privacy). In STRAP we combine light-weight goal-oriented analysis with heuristics to provide a simple yet effective design framework. We validate our analysis by demonstrating in a series of design experiments that STRAP is more efficient and effective than any one of the existing design frameworks, and provide quantitative and qualitative evidence of the value of using such frameworks as part of the design process.