Restless bandit problems assume time-varying reward distributions of the arms, which adds flexibility to the model but makes the analysis more challenging. We study learning algorithms over the unknown reward distributions and prove a sub-linear, O( √ T log T ), regret bound for a variant of Thompson sampling. Our analysis applies in the infinite time horizon setting, resolving the open question raised by Jung and Tewari (2019) whose analysis is limited to the episodic case. We adopt their policy mapping framework, which allows our algorithm to be efficient and simultaneously keeps the regret meaningful. Our algorithm adapts the TSDE algorithm of Ouyang et al. ( 2017) in a non-trivial manner to account for the special structure of restless bandits. We test our algorithm on a simulated dynamic channel access problem with several policy mappings, and the empirical regrets agree with the theoretical bound regardless of the choice of the policy mapping.

ObjectiveTo compare the intraoperative, radiological, and clinical short‐term outcomes of cementless total hip arthroplasties (THA) using a short stem (SS) and a conventional femoral stem (CS) in a randomized prospective control study.MethodsFrom June 2011 to October 2017, patients who underwent cementless THA for idiopathic osteonecrosis of the femoral head were recruited. Patients had a minimum 2 years of follow‐up after the operation. The patients were divided into two groups: those who underwent THA using an SS and those who underwent THA using a CS. SS were used in 34 patients (41 hips) and CS were used in 41 patients (45 hips). In both groups, the same cup was used in all cases, and the mean follow‐up periods were 63 (26–101) months in the SS and 64 (26–101) months in the CS groups. Intraoperative, clinical, and radiological evaluations were performed for the two groups.ResultsThere was no difference in the demographics of the two groups. There was one patient with a proximal femoral crack in the SS group and one with a distal femoral crack in the CS group. Clinically, the mean Harris hip score was improved in both groups at 2‐year follow‐up. Radiographically endosteal osseointegrations were found in 40 of 41 cases in the SS group and in 44 of 45 cases in the CS group. There was one case of dislocation in each group. In the SS group, the acetabular cup was changed and repositioned 7 months after the initial operation. Stem loosening, infection, ceramic breakage, and varus/valgus change were not observed. There was a statistically significant lower stress shielding effect in the SS group. There were no differences in vertical/parallel offset and leg length discrepancy.ConclusionThe intraoperative, radiological, and clinical evaluations in both groups showed good outcomes and there was no statistically significant difference between the two groups.

Annual average precipitation of Korea is 1,277 mm and around 2/3 of annual precipitation and 74 % of available water resources occurred during monsoon period. In recent years, many agricultural reservoirs have been heightened to increase flood control capacity, reduce flooding damage at downstream areas, and provide sustainable environmental flow during drought period. Thus in this study, the flood control effects of heightening of reservoir banks were simulated with HEC-ResSim and HEC-RAS models. These modes were applied to Bonghak reservoir and it was found that flood control effects were 3～4.5 % with 7 -m heightening. Also, with proper operation (1 m lower of full water level) of reservoir right before the monsoon period, flooding at downstream could be prevented even with design storm of 80 -year recurrence interval. As shown in this study, heightening of agricultural reservoir provides positive effects in flood control and flood damage reduction.

Four actual cultivations were prepared and a variety of soybean was cultivated. A H-flume, an automatic water level gauge and an automatic water sampler were installed at the outlet of each plot equipped for the measurement of flow rate and its water quality. The amount of rainfall of the study area in 2013 was measured as 975.6 mm which was much lower than the annual average rainfall of 1,271.8 mm, resulting in less occurrences in rainfall-runoff events. Rainfall-runoff events were occurred three times during the rainfall event of 4~5 July, 23 and 24 August. The characteristics of NPS pollution discharge of the plots and the reduction effect of the selected BMPs were analyzed during these events. The reduction effect of straw mat and soil amendments (Polyacrylamide (PAM) and Gypsum) on runoff ratio ranged between 38.2 and 92.9% (average 71.6%). The NPS pollution load reduced between 27.7 and 95.1% (average 70.0%) by the application of rice straw mat and soil conditioner when compared with that of control plot. Soybean yield (2,133.3 kg/ha) of the straw mat covered plots increased by 14.3% when compared with control (1,866.7 kg/ha). The effect of straw mat on the yield was not economically viable if the material and accompanying labor costs were considered. The data collected and analyzed on different soil textures and crops in this study are expected to be a fundamental reference for the expansion of the results to the application nationwide and the development of NPS pollution management policies.

Streamflow is composed of baseflow and direct runoff. However, most of streamflow during dry seasons depends on baseflow. Thus, baseflow analysis is very important to simulate streamflow of dry seasons. Generally, baseflow analysis is conducted using watershed-scale runoff models due to diffilculty of measuring baseflow. However, it is needed to understand and review how the model simulates baseflow because each model uses inherent baseflow analysis techniques. In this study, SWAT, HSPF, PRMS-IV were reviewed focusing on baseflow and soil water. HSPF and PRMS-IV calculate baseflow using the variables which depends on user, so the baseflow analysis results of HSPF and PRMS-IV are not consistent. Moreover, soil structures which were assumed from HSPF and PRMS-IV, since these two models assume soil structure as two soil zones and three conceptual reservoirs, were not enough to describe real soil structure. On the other hand, baseflow in SWAT is calculated using baseflow recession constant which can consider the characteristics of aquifer and also, soil structure in SWAT is similar to real soil structures. Thus, baseflow analysis result from SWAT was concluded as the most suitable and reliable model because SWAT can reflect the characteristics and soil structure which is close to reality.