No description available.

No description available.

Denna studie görs inom det arbetslivspedagogiska fältet och behandlar arbetsrotation på ICA-butiker. Syftet är att undersöka hur arbetsrotation sker hos medarbetare inom olika ICA-butiker och vilken betydelse det får för lärandet. Lärandet som undersöks är hur individer kan tillgodogöra sig yrkeskunskap utifrån de läroprocesser och lärandesituationer som uppstår i samband med arbetsrotation. De teoretiska verktyg som används grundar sig i Ellströms feed-forward och feed-back faser, de tre regleringsnivåerna av lärande samt en teori om subjektiva och objektiva handlingsutrymmen. Studien använder sig även av Börnfelts teori om en formell och informell organisation. Den metod som använts är kvalitativ och grundar sig i öppna intervjuer med åtta anställda från ICA-butiker med olika butiksformat, med det syftar vi till ICAs butiksformat; Nära, Supermarket, Kvantum och Maxi. Intervjuerna har transkriberats och analyserats ner till fyra olika teman; Arbetets innehåll, Avvikande situationer, Formellt Medarbetarskap och Informella Relationer. Resultatet visar att om anställda roterar mellan avdelningar eller ställs inför utmaning öppnar det upp individers handlingsutrymme, vilket leder till feed-forward faser som i sig genererar lärosituationer där anställda kan lära sig på en begreppslig- och intellektuell nivå. I flertal butiker bestämmer den formella organisationen det objektiva handlingsutrymmet men studien visar att den informella organisationen kan skapa sitt eget objektiva handlingsutrymme. Detta ställer visserligen krav på att det förekommer ett subjektivt handlingsutrymme i form av kunskap eller erfarenhet. I de butiker där ett tydligt objektivt handlingsutrymme ges tenderar anställda att hjälpa varandra inom och över avdelningar i högre utsträckning. I studien konstateras också att arbetsrotation gynnas av kollegiala relationer. Resultatet av studien kan användas för att vidare studera arbetsrotationens inverkan på lärande. Studien kan med fördel användas av yrkesverksamma inom ICA-butiker i syfte att få ökad förståelse för vilka konsekvenser som kan förbindas med införandet av arbetsrotation. / This paper examines job-rotation in ICA-stores within the pedagogy of working life. The purpose of this study is to determine how job-rotation works in ICA-stores and its relevance to learning from a employee perspective. The area of learning that we are investigating is how individuals assimilate knowledge from their professions through learning processes and situations during a job-rotation. The theoretical tools used in this study are based on Ellströms views on the feed-forward and feed-back phases, on Ellströms three levels of learning to conceptualise the act of learning, and finally on the theory of the subjective and objective spaces of action in regards to the learning situation. The paper mentions Börnfelts theories of formal and informal organisation as well. We used a qualitative method based on open interviews with eight employees from different ICA-stores, varied in both location and size. All interviews have been transcribed and analysed into four themes: The Content of the Work, Daily Obstacles/everyday situation, Formal Teamwork, and Informal Relations. The results of these interviews show that if employees rotate between departments, so called “revolving work tasks,” or face challenges within their own work tasks, the individuals acting space is opened up. This results in feed-forward and feed-back phases, which in turn generate the learning situations where the employees can learn on a conceptual and intellectual level. In several stores, the formal organisation creates an objective space of action for the employees, but the study shows that an informal organisation can also create their own objective space of action. If this is going to happen, employees must acquire a subjective space of action in the form of knowledge or experience. In a store with a clear objective space of action, employees tend to help each other more within and over departments. The study also states that a good relation between coworkers facilitates job-rotation. The results of the study can be used as a basis for further study of the way job-rotation affects learning. The study lends itself well to be used by people now working at ICA-stores, in order to get a better understanding of the consequences that can be connected to the introduction of job-rotation.

Streamed data classification problems (SDCPs) require classifiers with the ability to learn and to adjust to the underlying relationships in data streams, in real-time. This requirement poses a challenge to classifiers, because the learning task is no longer just to find the optimal decision boundaries, but also to track changes in the decision boundaries as new training data is received. The challenge is due to concept drift, i.e. the changing of decision boundaries over time. Changes include disappearing, appearing, or shifting decision boundaries. This thesis proposes an online learning approach for feed forward neural networks (FFNNs) that meets the requirements of SDCPs. The approach uses regularisation to optimise the architecture via the weights, and quantum particle swarm optimisation (QPSO) to dynamically adjust the weights. The learning approach is applied to a FFNN, which uses rectified linear activation functions, to form a novel SDCP classifier. The classifier is empirically investigated on several SDCPs. Both weight decay (WD) and weight elimination (WE) are investigated as regularisers. Empirical results show that using QPSO with no regularisation, causes the classifier to completely saturate. However, using QPSO with regularisation enables the classifier to dynamically adapt both its implicit architecture and weights as decision boundaries change. Furthermore, the results favour WE over WD as a regulariser for QPSO. / Dissertation (MSc)--University of Pretoria, 2020. / National Research Foundation (NRF) / Computer Science / MSc / Unrestricted

This thesis aims to bring clarity to the dimensioning aspects and limiting factors of the modular multilevel converter (MMC). Special consideration is given to the dc capacitors in the submodules as they are a driving factor for the size and weight of the converter. It is found that if the capacitor voltages are allowed to increase by 10% the stored energy must be 21 kJ/MW in order to compensate the capacitor voltage ripple. The maximum possible output power can, however, be increased by injecting a second-order harmonic in the circulating current. A great advantage of cascaded converters is the possibility to achieve excellent harmonic performance at low switching frequencies. Therefore, this thesis also considers the relation between switching harmonics, capacitor voltage ripple, and arm quantities. It is shown that despite subharmonics in the capacitor voltages, it is still possible to achieve periodic arm quantities. The balancing of the capacitor voltages is also considered in further detail. It is found that it is possible to balance the capacitor voltages even at fundamental switching frequency although this will lead to a comparably large capacitor voltage ripple. Therefore, in order to limit the peak-to-peak voltage ripple, it is shown that a predictive algorithm can be used in which the resulting switching frequency is approximately 2–3 times the fundamental frequency. This thesis also presents two new submodule concepts. The first submodule simply improves the trade-off between the switching frequency and capacitor voltage balancing. The second submodule includes the possibility to insert negative voltages which allows higher modulation indices compared to half-bridge submodules. A brief comparison of cascaded converters for ac-ac applications is also presented. It is concluded that the MMC appears to be well suited for ac-ac applications where input and output frequencies are close or equal, such as in interconnection of ac grids. In low-frequency applications such as low-speed drives, however, the difficulties with handling the energy variations in the converter arms are much more severe in the MMC compared to the other considered topologies. / <p>QC 20141010</p>

The aim of this study is to investigate how teachers say they give and perceive forward-looking feedback (feedforward) on pupils’ texts in grades three and four. To collect material for this purpose, semi-structured interviews were conducted with six teachers in the early years of compulsory school. The material from the interviews was analysed with the help of the mid-range coding method. The analysis was based on the two theoretical premises for the study, the sociocultural perspective and the formative attitude. The result shows that the teachers use several methods and devices for forward-looking feedback on the pupils’ process and on their first draft. These included questioning, assignment, checklist and digital tools. The study also revealed three central aspects of the teachers’ perceptions of their forward-looking feedback: shortage of time, organizational conditions and the pupils’ maturity. In view of the current lacuna in research on forward-looking feedback in the early years of compulsory school, the results of the study can contribute to inspiring, in-depth and effective knowledge and tools for practising forward-looking feedback.

Spectroscopic analysis has several applications in physics, chemistry, bioinformatics, geophysics, astronomy, etc. It has been widely used for detecting mineral samples, gas emission, and food volatiles. Machine learning algorithms for spectroscopic analysis focus on either regression or single-label classification problems. Using multi-label classification to identify multiple chemical components from the spectrum, has not been explored. In this thesis, we implement Feed-forward Neural Network with Optimal Thresholding (FNN-OT) identifying gas species among a multi gas mixture in a cluttered environment. Spectrum signals are initially processed by a feed-forward neural network (FNN) model, which produces individual prediction scores for each gas. These scores will be the input of a following optimal thresholding (OT) system. Predictions of each gas component in one testing sample will be made by comparing its output score from FNN against a threshold from the OT system. If its output score is larger than the threshold, the prediction is 1 and 0 otherwise, representing the existence/non-existence of that gas component in the spectrum. Using infrared absorption spectroscopy and tested on synthesized spectral datasets, our approach outperforms FNN itself and conventional binary relevance - Partial Least Squares with Binary Relevance (PLS-BR). All three models are trained and tested on 18 synthesized datasets with 6 levels of \signal-to-noise ratio and 3 types of gas correlation. They are evaluated and compared with micro, macro and sample averaged precision, recall and F1 score. For mutually independent and randomly correlated gas data, FNN-OT yields better performance than FNN itself or the conventional PLS-BR, by significantly by increasing recall without sacrificing much precision. For positively correlated gas data, FNN-OT performs better in capturing information of positive label correlation from noisy datasets than the other two models. / Graduate

This thesis concerns a longitudinal study of the practice of Live Peer Assessment on two University courses in Computer Science. By Live Peer Assessment I mean a practice of whole-class collective marking using electronic devices of student artefacts demonstrated in a class or lecture theatre with instantaneous aggregated results displayed on screen immediately after each grading decision. This is radically different from historical peer-assessment in universities which has primarily been asynchronous process of marking of students' work by small subsets of the cohort (e.g. 1 student artefact is marked by < 3 fellow students). Live Peer Assessment takes place in public, is marked by (as far as practically possible) the whole cohort, and results are instantaneous. This study observes this practice, first on a level 4 course in E-Media Design where students' main assignment is a multimedia CV (or resume) and secondly on a level 7 course in Multimedia Specification Design and Production where students produce a multimedia information artefact in both prototype and final versions. In both cases, students learned about these assignments from reviewing works done by previous students in Live Peer Evaluation events where they were asked to collectively publicly mark those works according to the same rubrics that the tutors would be using. In this level 4 course, this was used to help students get a better understanding of the marks criteria. In the level 7 course, this goal was also pursued, but was also used for the peer marking of students' own work. Among the major findings of this study are: • In the level 4 course student attainment in the final assessment improved on average by 13% over 4 iterations of the course, with very marked increase among students in the lower percentiles • The effectiveness of Live Peer Assessment in improving student work comes from o Raising the profile of the marking rubric o Establishing a repertoire of example work o Modelling the 'noticing' of salient features (of quality or defect) enabling students to self-monitor more effectively • In the major accepted measure of peer-assessment reliability (correlation between student awarded marks and tutor awarded marks) Live Peer Assessment is superior to traditional peer assessment. That is to say, students mark more like tutors when using Live Peer Assessment • In the second major measure (effect-size) which calculates if students are more strict or generous than tutors, (where the ideal would be no difference), Live Peer Assessment is broadly comparable with traditional peer assessment but this is susceptible to the conditions under which it takes place • The reason for the better greater alignment of student and tutor marks comes from the training sessions but also from the public nature of the marking where individuals can compare their marking practice with that of the rest of the class on a criterion by criterion basis • New measures proposed in this thesis to measure the health of peer assessment events comprise: Krippendorf's Alpha, Magin's Reciprocity Matrix, the median pairwise tutor student marks correlation, the Skewness and Kurtosis of the distribution of pairwise tutor student marking correlations • Recommendations for practice comprise that: o summative peer assessment should not take place under conditions of anonymity but that very light conditions of marking competence should be enforced on student markers (e.g. > 0.2 correlation between individual student marking and that of tutors) o That rubrics can be more suggestive and colloquial in the conditions of Live Peer Assessment because the marking criteria can be instantiated in specific examples of student attainment and therefore the criteria may be less legalistically drafted because a more holistic understanding of quality can be communicated.

There is a need for ecological modelling to help understand the dynamics in ecological systems, and thus aid management decisions to maintain or improve the quality of the ecological systems. This research focuses on non linear statistical modelling of observations from an estuarine system, Gippsland Lakes, on the south-eastern coast of Australia. Feed forward neural networks are used to model chlorophyll time series from a fixed monitoring station at Point King. The research proposes a systematic approach to modelling in ecology using feed forward neural networks, to ensure: (a) that results are reliable, (b) to improve the understanding of dynamics in the ecological system, and (c) to obtain a prediction, if possible. An objective filtering algorithm to enable modelling is presented. Sensitivity analysis techniques are compared to select the most appropriate technique for ecological models. The research generated a chronological profile of relationships between biophysical parameters and chlorophyll level for different seasons. A sensitivity analysis of the models was used to understand how the significance of the biophysical parameters changes as the time difference between the input and predicted value changes. The results show that filtering improves modelling without introducing any noticeable bias. Partial derivative method is found to be the most appropriate technique for sensitivity analysis of ecological feed forward neural networks models. Feed forward neural networks show potential for prediction when modelled on an appropriate time series. Feed forward neural networks also show capability to increase understanding of the ecological environment. In this research, it can be seen that vertical gradient and temperature are important for chlorophyll levels at Point King at time scales from a few hours to a few days. The importance of chlorophyll level at any time to chlorophyll levels in the future reduces as the time difference between them increases.

La durée d'analyse d'une scène visuelle complexe constitue une contrainte biologique fondamentale pour les modèles théoriques du traitement neuronal de l'information visuelle. La mesure d'indices psychophysiques et électrophysiologiques lors d'une tâche de catégorisation visuelle rapide permet de démontrer que le système visuel humain transforme une scène visuelle complexe en une représentation de haut niveau, portant une valeur sémantique ou comportementale, en seulement 150 ms. Ce résultat est indépendant du caractère naturel ou artificiel de la catégorie cible utilisée. L'activité cérébrale enregistrée avant 150 ms reflète les propriétés physiques du stimulus, et n'est pas corrélée avec le comportement du sujet. D'un point de vue théorique, quels mécanismes neuronaux peuvent sous-tendre une telle capacité?<br />L'architecture hiérarchique du système visuel des primates et les délais de conduction électrique suggèrent que l'information visuelle circule majoritairement vers l'avant, et qu' un seul potentiel d'action par neurone pourra coder l'information entre 2 étapes successives. Ceci exclut le codage par fréquence d'émission de potentiels d'action, classiquement utilisé par les modèles de la vision. J'illustre cependant le fait que l'information visuelle peut être encodée par les dates relatives d'émission des potentiels d'action sur une population neuronale, l'information la plus saillante étant toujours représentée par les premières décharges. La structure temporelle de la première vague de potentiels d'action générée par la rétine en réponse à une stimulation visuelle peut porter explicitement l'information. Cette vague se propageant à travers le système est régénérée à chaque étape de traitement, et sa structure temporelle peut être modifiée par (i) la sélectivité des neurones corticaux, (ii) des interactions latérales et (iii) des influences attentionnelles provenant d'aires cérébrales centrales.