56016 research outputs found
Sort by
Heavy-tailed Linear Bandit with Huber Regression
No full textLinear bandit algorithms have been extensively studied and have shown successful in sequential decision tasks despite their simplicity. Many algorithms however work under the assumption that the reward is the sum of linear function of observed contexts and a sub-Gaussian error. In practical applications, errors can be heavy-tailed, especially in financial data. In such reward environments, algorithms designed for sub-Gaussian error may underexplore, resulting in suboptimal regret. In this paper, we relax the reward assumption and propose a novel linear bandit algorithm which works well under heavy-tailed errors as well. The proposed algorithm utilizes Huber regression. When contexts are stochastic with positive definite covariance matrix and the (1 + ??)-th moment of the error is bounded by a constant, we show that the high-probability upper bound of the regret is O(???dT 1+ 1 ?? (log dT) 1+ ?? ?? ), where d is the dimension of context variables, T is the time horizon, and ?? ??? (0, 1]. This bound improves on the state-of-the-art regret bound of the Median of Means and Truncation algorithm by a factor of ???log T and ???d for the case where the time horizon T is unknown. We also remark that when ?? = 1, the order is the same as the regret bound of linear bandit algorithms designed for sub-Gaussian errors. We support our theoretical findings with synthetic experiments
Evaluations of patient-specific bolus fabricated by mold-and-cast method using computer numerical control machine tools
No full textThe patient-specific bolus fabricated by a mold-and-cast method using a 3D printer (3DP) and silicon rubber has been adopted in clinical practices. Manufacturing a mold using 3DP, however, can cause time delays due to failures during the 3D printing process. Thereby, we investigated an alternative method of the mold fabrication using computer numerical control (CNC) machine tools. Treatment plans were conducted concerning a keloid scar formed on the ear and nose. The bolus structures were determined in a treatment planning system (TPS), and the molds were fabricated using the same structure file but with 3DP and CNC independently. Boluses were then manufactured using each mold with silicone rubbers. We compared the geometrical difference between the boluses and the planned structure using computed tomography (CT) images of the boluses. In addition, dosimetric differences between the two measurements using each bolus and the differences between the measured and calculated dose from TPS were evaluated using an anthropomorphic head phantom. Geometrically, the CT images of the boluses fabricated by the 3DP mold and the CNC mold showed differences compared to the planned structure within 2.6 mm of Hausdorff distance. The relative dose difference between the measurements using either bolus was within 2.3%. In conclusion, the bolus made by the CNC mold benefits from a stable fabricating process, retaining the performance of the bolus made by the 3DP mold
Semi-analytic shooting methods for Burgers equation
No full textWe implement new semi-analytic shooting methods for the stationary viscous Burgers' equation by modifying the classical time differencing methods. When the viscosity is small, a very stiff boundary layer appears and this boundary layer causes significant difficulties to approximate the solution for Burgers' equation. To overcome this issue and improve the numerical quality of the shooting methods with the classical Integrating Factor (IF) methods and Exponential Time Differencing (ETD) methods, we first employ the singular perturbation analysis for Burgers' equation, and derived the so-called correctors that approximate the stiff part of the solution. Then, we build our new semianalytic shooting methods for the stationary viscous Burgers' equation by embedding these correctors into the IF and ETD methods. By performing numerical simulations, we verify that our new schemes, enriched with the correctors, give much better approximations, compared with the classical schemes.(c) 2022 Elsevier B.V. All rights reserved
Managerial perspectives on climate change and stock price crash risk
Full text linkIn this study, we examine the effects of manager's perspectives on climate change on stock price crash risk. The analysis confirms that manager's climate change perspective is negatively associated with future stock price crash risk likelihood. Various channel tests show that investor attention and analyst coverage are potential channels through which a firm's climate change perspective improves financial stability and ultimately reduces crash risk. Our results are also robust to alternative climate change perspective measures
????????? ????????? ??????????????? ???????????? ????????? LSTM??? ????????? SIRD ??????
No full textDepartment of Mathematical SciencesFrom the end of 2019, COVID-19 occurred by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) in Wuhan, China. As the number of confirmed cases worldwide has increased exponentially, the World Health Organization (WHO) declared COVID-19 a pandemic on March 11, 2020. As of August 8, 2020, there are 14,562 confirmed cases and 304 deaths in Korea. As the number of covid 19 confirmed cases increases, economic and cultural losses increase, so it is important to predict the number of confirmed cases well and establish policies. As the number of infected cases increases, covid 19 search query volume also increases, hence this will be applied to predicting the number of infected cases. The parameters: infection rate, recovery rate, and death rate were obtained using the SIRD equation. Among them, the infection rate and covid 19 search query volume data are combined to make an input value of LSTM suitable for time series prediction. Using the output obtained using LSTM as a new infection rate, substitute it into the SIRD ode equation to obtain infected data. Comparing the case when the query was added and the case when the query was not added, unlike expected the case without the query predicted the infected data better. In the future, rather than adding web data to parameter values, we will try to pursue the direction of transforming the SIRD model itself so that wed data can be included.ope
An autopsy study of hollow fiber and multibore ultrafiltration membranes from a pilot-scale ultra high-recovery filtration system for surface water treatment
No full textThe organic fouling characteristics of hollow fiber ultrafiltration (HFUF) and multibore ultrafiltration (MBUF) mem-branes from long-term ultrafiltration (UF) membrane systems were systemically investigated in this study. The objec-tive was to obtain insights into the fouling behavior of dissolved organic matter (DOM) in a pilot-scale ultra-high-recovery membrane filtration system (p-UHMS) used for surface water treatment. The pilot system consisted of a series of two different UF membranes (1st stage: polyvinylidene fluoride (PVDF) HFUF and 2nd stage: polyethersulfone (PES) MBUF). It was designed to feed the HFUF concentrate to the MBUF membranes to achieve >= 99.5 % total water recov-ery for surface water treatment, as these advances might enhance the production efficiencies of drinking water. The experimental results confirmed that hydrophobic DOM controlled the formation of HFUF membrane organic fouling, whereas hydrophilic DOM, including polysaccharide-like and protein-like matter, promoted MBUF membrane fouling. These opposing trends were attributed to the hydrophilic characteristics of the MBUF membrane surfaces (contact angle: PVDF = 90-130 degrees and PES <= 80 degrees), which reduced the hydrophobic interactions between the UF membrane sur-faces and foulants. The performance declines of the MBUF membrane due to fouling layer formation was considerably severer than those of the HFUF membrane, decreasing total permeate water in the p-UHMS. Moreover, the quantity of the desorbed MBUF membrane foulants via 0.1 N NaOH was roughly 7.2 times larger than that of the desorbed HFUF membrane foulants through 0.1 N NaOH, indicating that alkaline-based cleaning agent could much more efficiently recover the performance of the fouled MBUF membranes. Hence, adequate cleaning strategies using alkaline-based agent for the MBUF membrane appeared to be essential for preventing the performance deterioration of the p-UHMS
Artificial neural network-based wall-modeled large-eddy simulations of turbulent channel and separated boundary layer flows
No full textWall-models in a large-eddy simulation (LES) are essential to alleviate the large near-wall resolution requirements for high-Reynolds-number turbulent flow simulations. Among the existing wall-models for a LES, an equilibrium wall-stress model has the highest computational efficiency. Because this model has limitations, such as a lack of non-equilibrium effects and the assumption of a particular law of the wall in the mean velocity, we propose artificial neural network-based wall-stress models (AWMs). The input variables for the AWMs are extracted from the decomposition of the skin-friction coefficient proposed by Fukagata et al. [1], and the AWMs are shown to be able to predict the wall-shear stress in complex flows accurately. The performance of the AWMs is tested for two types of flows, a fully developed turbulent channel flow and a separated turbulent boundary layer flow. A direct comparison of the turbulence statistics with those obtained by previous wall-models (i.e., a log-law-based wall-stress model and a non-equilibrium wall-stress model) shows that better predictions are achieved using the AWMs for both flows, even with untrained Reynolds numbers. When using a coarse grid along the wall-normal direction in wall-modeled LESs (WMLESs) with the AWMs, an upward shift of the mean velocity profile (positive log-layer mismatch, LLM) compared to direct numerical simulation data is found, consistent with previous studies. However, this LLM problem can be overcome by imposing a filtered wall-normal velocity at the wall that is dynamically determined based on the continuity equation and the Taylor series expansion within wall-adjacent cells
Stretchable skin hydration sensor based on hygroscopic and ion conductive polymer composites
No full textTransepidermal water loss (TEWL) is water emission from the skin induced by the diffusion of water molecules. Because TEWL is sensitive to damage on the outermost skin layer, it has been utilized as an indicator of skin barrier function. Here, we demonstrate a stretchable hydration sensor to monitor TEWL based on a porous thermoplastic polyurethane (TPU) film coated with ion conductive polyvinyl alcohol (PVA)/lithium chloride (LiCl) composite layers. The stretchable hydration sensor with the large surface area and the hygroscopic PVA/ LiCl layer coated on the porous structure exhibits a remarkable relative current change (Delta I/I0 (%) similar to 107) under a wide humidity range (10-95 % relative humidity (RH)). Moreover, the highly deformable characteristics of the device up to 50 % strain makes it an appropriate tool for detecting water emission from a curved skin surface. For a proof-of-concept demonstration of monitoring skin barrier function, the hydration sensor can precisely perceive the change in TEWL in response to various external treatments (heat, damage, oil treatment, and applying cosmetics). Finally, the stretchable hydration sensor array enables non-contact finger motion perception through water loss detection, making it suitable for use in touchless human-machine interfaces
Delegation of Information Acquisition, Information Asymmetry, and Outside Option
No full textWe study how a principal with an outside option optimally delegates information acquisition to an agent in a parsimonious environment in which the principal can observe neither the agent's effort nor signal realizations. When the principal chooses an outside option, the true state is not revealed and thus not contractible. We precisely characterize an optimal contract for the principal, illustrating how to construct an optimal contract
Multiferroicity of 2H-BaMnO3 Single Crystal
No full textGeometric ferroelectrics are called as improper ferroelectrics where geometric structural constraints, rather than typical cation-anion pairing, induce proper ferroelectric polarization. Among the geometric structural distortions, trimerization of MnO5 bipyramids is observed in RMnO3 (R = Ho, Er, Yb, Lu, Y), which share same symmetric group with 2H-BaMnO3. And the distortion induces intriguing multiferroicity such as topological vortex-antivortex, angle dependent conducting domain walls, enhanced magnetoelectric coupling at domain walls, etc. In hexagonal RMnO3, it has been unavailable to study thermodynamic behavior of trimerization and polarization because of high ferroelectric transition temperature (1120-1435??C). But, it was reported that polycrystalline 2H-BaMnO3 exhibits antiferromagnetic order at 59 K [1], and ferroelectric transition at 130 K [2]. Here, we present ferroelectric and magnetic properties of single crystalline 2H-BaMnO3.
[1] E. J. Cussen and P. D. Battle, Chem. Mater. 12, 831-838 (2000)
[2] Stanislav Kamba et al, Physical Review B 95, 174103 (2017