abstract: Visual object recognition has achieved great success with advancements in deep learning technologies. Notably, the existing recognition models have gained human-level performance on many of the recognition tasks. However, these models are data hungry, and their performance is constrained by the amount of training data. Inspired by the human ability to recognize object categories based on textual descriptions of objects and previous visual knowledge, the research community has extensively pursued the area of zero-shot learning. In this area of research, machine vision models are trained to recognize object categories that are not observed during the training process. Zero-shot learning models leverage textual information to transfer visual knowledge from seen object categories in order to recognize unseen object categories. Generative models have recently gained popularity as they synthesize unseen visual features and convert zero-shot learning into a classical supervised learning problem. These generative models are trained using seen classes and are expected to implicitly transfer the knowledge from seen to unseen classes. However, their performance is stymied by overfitting towards seen classes, which leads to substandard performance in generalized zero-shot learning. To address this concern, this dissertation proposes a novel generative model that leverages the semantic relationship between seen and unseen categories and explicitly performs knowledge transfer from seen categories to unseen categories. Experiments were conducted on several benchmark datasets to demonstrate the efficacy of the proposed model for both zero-shot learning and generalized zero-shot learning. The dissertation also provides a unique Student-Teacher based generative model for zero-shot learning and concludes with future research directions in this area. / Dissertation/Thesis / Masters Thesis Computer Science 2020

This dissertation presents a study on Grade 10 learners’ understanding of key Mathematics concepts in selected secondary schools in Chibombo District of Zambia. The performance of learners in Mathematics countrywide has not been gratifying. This study therefore sought to investigate the Grade 10 learners’ understanding of key Mathematics concepts. This poor performance in Mathematics by learners in many schools at grade 12 final examination in Chibombo District prompted the researcher to conduct this study. The target population was 250 Grade 10 learners in five secondary schools in Chibombo District. The data were collected by means of questionnaire and focus group discussions. A concurrent triangulation design was adopted for the study in which the researcher simultaneously gathered both quantitative and qualitative data merged it using both quantitative and qualitative data analysis. The study revealed that Grade 10 learners’ understanding of the key Mathematics concepts of zero, equal, function, and variable is unsatisfactory. However, it is surprising how these learners can reach secondary school level with such a weak understanding of mathematical concepts, which are ubiquitous in the subject. It is anticipated that the findings of this study will provide curriculum developers with new insights into this problem so that they come up with a curriculum that focuses on learners’ understanding of basic concepts that serve as building blocks to next levels. The Ministry of General Education should rigorously monitor this process and provide both technical and financial support required. Finally, based on the findings of this study, the researcher gave some recommendations for improving learners’ understanding of mathematical concepts in schools. / Mathematics Education / M. Ed. (Mathematics Education)

abstract: Despite the concern over school shootings in the last twenty years, little has been done to prevent these events. This paper addresses the need to have a consistent definition of a school shooting. The policies that have been enacted in the eight states with the most shootings in the past 20 are categorized and compared to the number of incidences and victims. The study concludes that states need fewer reactionary policies and more policies based on systematic research; these states pass a majority of zero-tolerance, which are shown to be ineffective in preventing school shootings suggesting a need for a new approach to writing and addressing policy. / Dissertation/Thesis / Masters Thesis Integrative Social Science 2020

van der Waals heterostructures meet other low-dimensional materials. Stacking of about 1 nm thick nanosheets with out-of-plane anchor groups functionalized with fullerenes integrates this zero-dimensional material into layered heterostructures with a well-defined chemical composition and without degrading the mechanical properties. The developed modular and highly applicable approach enables the incorporation of other low-dimensional materials, e.g. nanoparticles or nanotubes, into heterostructures significantly extending the possible building blocks.

One way to improve dependency parsing scores for low-resource languages is to make use of existing resources from other closely related or otherwise similar languages. In this paper, we look at eleven Uralic target languages (Estonian, Finnish, Hungarian, Karelian, Livvi, Komi Zyrian, Komi Permyak, Moksha, Erzya, North Sámi, and Skolt Sámi) with treebanks of varying sizes and select transfer languages based on geographical, genealogical, and syntactic distances. We focus primarily on the performance of parser models trained on various combinations of geographically proximate and genealogically related transfer languages, in target-trained, zero-shot, and cross-lingual configurations. We find that models trained on combinations of geographically proximate and genealogically related transfer languages reach the highest LAS in most zero-shot models, while our highest-performing cross-lingual models were trained on genealogically related languages. We also find that cross-lingual models outperform zero-shot transfer models. We then select syntactically similar transfer languages for three target languages, and find a slight improvement in the case of Hungarian. We discuss the results and conclude with suggestions for possible future work.

Free-space optics (FSO) is an unlicensed communication technology that uses the free space as a propagation medium to connect two communicating terminal wire- lessly [1]. It is an attractive solution to the last-mile connectivity problems in commu- nication networks, mainly when installing optical fibers is expensive or unavailable. A possible idea to increase the throughput of wireless optical links in free space is to use spatial multiplexing (SMM) [2]. Optical beam distortion due to propagation through a turbulent channel is one of the main factors limiting performance of such a system. Therefore, overcoming the effect of turbulence is a major problem for structured light optical communication in free space. Usually, this problem is approached by using adaptive optics systems and various methods of digital signal processing (DSP) on the receiver side [3–5]. Recently, an idea of optical channel pre-compensation to mit- igate inter-modal crosstalk was proposed [6] and experimentally validated [7]. Such a method, if implemented, will allow the use of entirely passive receivers or, in the case of full-duplex transmission, increase throughput. Here, the performance of a zero-forcing precoding technique to mitigate the effects of an optical turbulence in a Laguerre Gaussian mode based SMM FSO is investigated. Equally, details on a close to reality simulation of the atmospheric turbulence and beam propagation are provided.

京都大学 / 0048 / 新制・課程博士 / 博士(理学) / 甲第21558号 / 理博第4465号 / 新制||理||1641(附属図書館) / 京都大学大学院理学研究科物理学・宇宙物理学専攻 / (主査)教授 大西 明, 教授 青木 慎也, 教授 田中 貴浩 / 学位規則第4条第1項該当 / Doctor of Science / Kyoto University / DFAM

京都大学 / 新制・課程博士 / 博士(情報学) / 甲第23317号 / 情博第753号 / 新制||情||129(附属図書館) / 京都大学大学院情報学研究科社会情報学専攻 / (主査)教授 神田 崇行, 教授 吉川 正俊, 教授 岡部 寿男 / 学位規則第4条第1項該当 / Doctor of Informatics / Kyoto University / DFAM

In the debate on how to present and prevent climate change, the opinions have been conflicted. Some argue that a pessimistic projection will trigger a doomsday-effect and others argue that an optimistic projection enhances the risk of remaining passive. However, there is a fine line between what is too optimistic and too pessimistic. With the basis from choice architecture and the Zero Contribution Thesis, this study has dealt with how, and if, different projections could influence people's attitudes through nudges. By conducting two surveys which project different outcomes of climate change, one pessimistic, and one optimistic, this study aims at contributing to a greater understanding of how the use of communication can influence peoples’ attitudes and commitment towards climate change. When being a choice architect and constructing a nudge, it is crucial to carefully consider the design of the nudge since it is vital to understand that small details have a significant effect on the choices a person makes. By making small changes in our everyday life, we can reduce the impact of our ecological footprint. With the results it can be concluded that a purely pessimistic or optimistic projection may not be the most effective method when trying to influence environmentally friendly behaviour. Therefore, it could be beneficial to use a combination of both an optimistic- and pessimistic- projection of climate change. Furthermore, nudges could be considered as a legitimate policy tool when trying to create environmentally friendly behaviour.

Proper disposal of industrial brine has been a critical environmental challenge. Driven by the even-tightening environment protection regulations, the Zero-Liquid-Discharge (ZLD) has gradually become mandatory option for brine disposal, but its application is limited by the intensive energy consumption. The recent development of solar crystallizer provides a new strategy to achieve ZLD brine disposal. However, the research on solar crystallizer, employing photothermal material to convert solar energy to heat for interfacial brine evaporation and crystallization, is still at the early stage. This dissertation thoroughly investigated the solar crystallizer-based ZLD technology in a broad scientific and application context. The scaling formation while treating real brine, which has been the major barrier to the application of solar crystallizer, was confirmed first with a solar crystallizer device. With a rational designed anti-scaling mechanism, the scaling-free crystallization behavior and stable high water evaporation rate of 2.42 kg m-2 h-1 was achieved when treating real seawater brine. After verifying the feasibility of solar crystallizer towards real brine treatment, its performance was further improved by integrating convective airflow, which provided considerable environmental energy for water evaporation. Both experiment results and COMSOL simulation results confirmed that the maximum environmental energy harvesting can be achieved with the proper size of solar crystallizer. At last, this dissertation pioneered a novel concept of integrating adsorption process into solar crystallizer for simultaneously ZLD brine treatment and potassium extraction. Owing to the special ion concentration behavior of solar crystallizer, the adsorption capacity and selectivity coefficient of absorbent was enhanced by 19.5% and 48.8%, respectively, comparing with traditional bulk adsorption. This dissertation potentially unlocks a new generation of ZLD technology with low carbon footprint and source recovery. More research efforts will be inspired on its applications in real scenarios.