Judgements of and sensitivity to style are ubiquitous. People become sensitive to the structural regularities of complex or “polymorphous” categories through exposure to individual examples, which allows them respond to new items that are of the same style as those previously experienced. This thesis investigates whether a dimension reduction mechanism could account for how people learn about the structure of complex categories. That is, whether through experience, people extract the primary dimensions of variation in a category and use these to analyse and categorise subsequent instances. We used Singular Value Decomposition (SVD) as the method of dimension reduction, which yields the main dimensions of variation of pixel-based stimuli (eigenvectors). We then tested whether a simple autoassociative network could learn to distinguish paintings by Picasso and Braque which were reconstructed from only these primary dimensions of variation. The network could correctly classify the stimuli, and its performance was optimal with reconstructions based on just the first few eigenvectors. Then we reconstructed the paintings using either just the first 10 (early reconstructions) or all 1,894 eigenvectors (full reconstructions), and asked human participants to categorise the images. We found that people could categorise the images with either the early or full reconstructions. Therefore, people could learn to distinguish category membership based on the reduced set of dimensions obtained from SVD. This suggests that a dimension reduction mechanism analogous to SVD may be operating when people learn about the structure and regularities in complex categories.

Judgements of and sensitivity to style are ubiquitous. People become sensitive to the structural regularities of complex or “polymorphous” categories through exposure to individual examples, which allows them respond to new items that are of the same style as those previously experienced. This thesis investigates whether a dimension reduction mechanism could account for how people learn about the structure of complex categories. That is, whether through experience, people extract the primary dimensions of variation in a category and use these to analyse and categorise subsequent instances. We used Singular Value Decomposition (SVD) as the method of dimension reduction, which yields the main dimensions of variation of pixel-based stimuli (eigenvectors). We then tested whether a simple autoassociative network could learn to distinguish paintings by Picasso and Braque which were reconstructed from only these primary dimensions of variation. The network could correctly classify the stimuli, and its performance was optimal with reconstructions based on just the first few eigenvectors. Then we reconstructed the paintings using either just the first 10 (early reconstructions) or all 1,894 eigenvectors (full reconstructions), and asked human participants to categorise the images. We found that people could categorise the images with either the early or full reconstructions. Therefore, people could learn to distinguish category membership based on the reduced set of dimensions obtained from SVD. This suggests that a dimension reduction mechanism analogous to SVD may be operating when people learn about the structure and regularities in complex categories.

京都大学 / 新制・課程博士 / 博士(情報学) / 甲第23323号 / 情博第759号 / 新制||情||129(附属図書館) / 京都大学大学院情報学研究科数理工学専攻 / (主査)教授 中村 佳正, 教授 矢ヶ崎 一幸, 准教授 辻本 諭 / 学位規則第4条第1項該当 / Doctor of Informatics / Kyoto University / DFAM

The Polar Decomposition (PD) of a dense matrix is an important operation in linear algebra, while being a building block for solving the Symmetric Eigenvalue Problem (SEP) and computing the Singular Value Decomposition (SVD). It can be directly calculated through the SVD itself, or iteratively using the QR Dynamically-Weighted Halley (QDWH) algorithm. The former is difficult to parallelize due to the preponderant number of memory-bound operations during the bidiagonal reduction. The latter is an iterative method, which performs more floating-point operations than the SVD approach, but exposes at the same time more parallelism. Looking at the roadmap of the hardware technology scaling, algorithms perform- ing floating-point operations on locally cached data should be favored over those requiring expensive horizontal data movement. In this context, this thesis investigates new high-performance algorithmic designs of QDWH algorithm to compute the PD. Originally introduced by Nakatsukasa et al. [1, 2], our algorithmic contributions include mixed precision techniques, task-based formulations, and parallel asynchronous executions. Moreover, by making the PD competitive, its application to the SEP and the SVD becomes practical. In particular, we introduce for the first time new algorithms for partial SVD decomposition using QDWH. By the same token, we extend the QDWH to support partial eigen decomposition for SEP. We present new high-performance implementations of the QDWH-based algorithms relying on fine-grained computations, which allows exploiting the sparsity of the underlying data structure. To demonstrate performance efficiency, portability and scalability, we conduct benchmarking campaigns on some of the latest shared/distributed-memory systems. Our QDWH-based algorithm implementations outperform the state-of-the-art numerical libraries by up to 2.8x and 12x on shared and distributed-memory, respectively. The task-based QDWH has been integrated into the Chameleon library (https://gitlab.inria.fr/solverstack/chameleon) for support on shared-memory systems with hardware accelerators. It is also currently being used by astronomers from the Subaru telescope located at the summit of Mauna Kea, Hawaii, USA. The distributed-memory software library of QDWH and its SVD extension are freely available under modified-BSD license at https: //github.com/ecrc/qdwh.git and https://github.com/ecrc/ksvd.git, respectively. Both software libraries have been integrated into the Cray Scientific numerical library LibSci v17.11.1 and v19.02.1.

Recommender System (RS) has become one of the most important component for many companies, such as YouTube and Amazon. A recommender system consists of a series of algorithms which predict and recommend products to users. This report covers the selection of many open source recommender system projects, and movie predictions are made using the selected recommender system. Based on the predictions, a comparison was made between precision and an improved precision algorithm. The selected RS uses singular value decomposition in the field of collaborative filtering. Based on the recommendation results produced by the RS, the comparison between precision and the improved precision algorithms showed that the result of improved precision is slightly higher than precision in different cutoff values and different dimensions of eigenvalues. / Rekommendationssystem har blivit en av de viktigaste beståndsdelar för många företag, såsom YouTube och Amazon. Ett rekommendationssystem består av en serie av algoritmer som förutsäger och rekommenderar produkter till användare. Denna rapport omfattar valet av många öppen källkod rekommendationssystem projekt, och filmprognoser görs med det valda rekommendationssystemet. Baserat på filmprognoser, gjordes en jämförelse mellan precision och en förbättrad precision algoritmer. Det valda rekommendationssystemet använder singulärvärdesuppdelning som kollaborativ filtrering. Baserat på rekommendationsresultat som produceras av rekommendationssystemet, jämförelsen mellan precision och den förbättrade precisions algoritmer visade att resultatet av förbättrad precision är något högre än precision i olika brytvärden och olika dimensioner av egenvärden.

Climate variability is important to understand as its effects on groundwater are complex than surface water. Climate association between Groundwater Storage (GWS) and sea level changes have been missing from the Intergovernmental Panel on Climate Change, demanding a requisite study of their linkage and responses. The current dissertation is primarily focused on the ongoing issues that have not been focused on the previous literatures. Firstly, the study evaluated the effects of short-term persistence and abrupt shifts in sea level records along the US coast by utilizing popular robust statistical techniques. Secondly, the study evaluated the variability in groundwater due to variability in hydroclimatic variables like sea surface temperature (SST), precipitation, sea level, and terrestrial water storage. Moreover, a lagged correlation was also analyzed to obtain their teleconnection patterns. Lastly, the relationship between the groundwater rise and one of the most common short-term climate variability, ENSO was obtained. To accomplish the research goals the current dissertation was subdivided into three research tasks.The first task attempted to answer a major question, Is sea level change affected by the presence of autocorrelation and abrupt shift? This question reflects the importance of trend and shift detection analysis in sea level. The primary factor driving the global sea level rise is often related to climate change. The current study investigates the changes in sea level along the US coast. The sea level records of 59 tide gauge data were used to evaluate the trend, shift, and persistence using non-parametric statistical tests. Mann-Kendall and Pettitt’s tests were utilized to estimate gradual trends and abrupt shifts, respectively. The study also assessed the presence of autocorrelation in sea level records and its effect on both trend and shift was examined along the US coast. The presence of short-term persistence was found in 57 stations and the trend significance of most stations was not changed at a 95% confidence level. Total of 25 stations showed increasing shift between 1990–2000 that was evaluated from annual sea level records. Results from the current study may contribute to understanding sea level variability across the contiguous US. The second task dealt with variability in the Hydrologic Unit Code—03 region. It is one of the major U.S. watersheds in the southeast in which most of the variability is caused by Sea Surface Temperature (SST) variability in the Pacific and Atlantic Ocean, was identified. Furthermore, the SST regions were identified to assess its relationship with GWS, sea level, precipitation, and terrestrial water storage. Temporal and spatial variability were obtained utilizing the singular value decomposition statistical method. A gridded GWS anomaly from the Gravity Recovery and Climate Experiment (GRACE) was used to understand the relationship with sea level and SST. The negative pockets of SST were negatively linked with GWS. The identification of teleconnections with groundwater may substantiate temporal patterns of groundwater variability. The results confirmed that the SST regions exhibited El Niño Southern Oscillation patterns, resulting in GWS changes. Moreover, a positive correlation between GWS and sea level was observed on the east coast in contrast to the southwestern United States. The findings highlight the importance of climate-driven changes in groundwater attributing changes in sea level. Therefore, SST could be a good predictor, possibly utilized for prior assessment of variabilities plus groundwater forecasting. The primary goal of the third task is to better understand the effects of ENSO climate patterns on GWS in the South Atlantic-Gulf region. Groundwater issues are complex and different studies focused on groundwater depletion while few emphasized, “groundwater rise”. The current research is designed to develop an outline for assessing how climate patterns can affect groundwater fluctuation, which might lead to groundwater rise. The study assessed the effect of ENSO phases on spatiotemporal variability of groundwater using Spearman Rank Correlation. A significant positive correlation between ENSO and GWS was observed. An increasing trend was detected in GWS where most grids were observed in Florida by utilizing the non-parametric Mann-Kendall. A positive magnitude of the trend was also detected by utilizing Theil-Sen’s Slope method with high magnitude in the mid-Florida region. The highest GWS anomalies were observed in the peak of El Niño events and the lowermost GWS was observed during La Niña events. Furthermore, most of the stations were above normal groundwater conditions. This study provides a better understanding of the research gap between groundwater rise and ENSO.

京都大学 / 0048 / 新制・課程博士 / 博士(情報学) / 甲第17968号 / 情博第512号 / 新制||情||91(附属図書館) / 30798 / 京都大学大学院情報学研究科数理工学専攻 / (主査)教授 中村 佳正, 教授 西村 直志, 准教授 山下 信雄 / 学位規則第4条第1項該当 / Doctor of Informatics / Kyoto University / DFAM

原著論文リスト[A1]: “The final publication is available at Springer via http://dx.doi.org/10.1007/s11075-012-9607-5.”. [A2]: “The final publication is available at Springer via http://dx.doi.org/10.1007/s10092-013-0085-5.”, [A3]: DOI“10.1016/j.camwa.2015.11.022” / 京都大学 / 0048 / 新制・課程博士 / 博士(情報学) / 甲第19859号 / 情博第610号 / 新制||情||106(附属図書館) / 32895 / 京都大学大学院情報学研究科数理工学専攻 / (主査)教授 中村 佳正, 教授 矢ｹ崎 一幸, 教授 山下 信雄 / 学位規則第4条第1項該当 / Doctor of Informatics / Kyoto University / DFAM

The current research investigates the use of solid-embedded thermocouples for determining accurate transient temperature measurements within a solid medium, with emphasis on measurements intended for use in inverse heat conduction problems. Metal casting experiments have been conducted to collect internal mold temperatures to be used, through inverse conduction methods, to estimate the heat exchange between a casting and mold. Inverse conduction methods require accurate temperature measurements for valid boundary estimates. Therefore, various sources of thermocouple measurement uncertainty are examined and some suggestions for uncertainty reduction are presented. Thermocouple installation induced bias uncertainties in experimental temperature data are dynamically corrected through the development and implementation of an embedded thermocouple correction (ETC) transfer function. Comparisons of experimental data to dynamically adjusted data, as well as the inverse conduction estimates for heat flux from each data set, are presented and discussed.

No description available.