Cheng-Chieh Peng

Drone delivery is swiftly on the rise. Different from most prior efforts centered on the last-mile problem where the drone flies from the source to the destination house but overlooks the landing part [1, 2]. In our full paper [3], we make the first attempt on the last-hundred-feet problem, which aims to land at a convenient and safe point (say D-point, in front of the door) in a fully autonomous manner. In this demo, we specifically focus on the key technical component Structural Semantic Segmentation (SSS), a novel semantic segmentation-based approach to identify D-point. Unlike the classical SS, it capitalizes the common single-family house (SFH) structures to refine the misclassified region from SS results. Our results show that the detection precision of boundary between the house and pavement (indicating the door location) reaches 84%, compared to 16% achieved by SS method, and only adds 430 ms to the processing time. We leverage SSS to progressively update the D-point guiding it closer to the door (Figure 1a). Video can be found at https://youtu.be/G6I9XzCyHFQ

In the assessment of car insurance claims, the claim rate for car insurance presents a highly skewed probability distribution, which is typically modeled using Tweedie distribution. The traditional approach to obtaining the Tweedie regression model involves training on a centralized dataset, when the data is provided by multiple parties, training a privacy-preserving Tweedie regression model without exchanging raw data becomes a challenge. To address this issue, this study introduces a novel vertical federated learning-based Tweedie regression algorithm for multi-party auto insurance rate setting in data silos. The algorithm can keep sensitive data locally and uses privacy-preserving techniques to achieve intersection operations between the two parties holding the data. After determining which entities are shared, the participants train the model locally using the shared entity data to obtain the local generalized linear model intermediate parameters. The homomorphic encryption algorithms are introduced to interact with and update the model intermediate parameters to collaboratively complete the joint training of the car insurance rate-setting model. Performance tests on two publicly available datasets show that the proposed federated Tweedie regression algorithm can effectively generate Tweedie regression models that leverage the value of data from both parties without exchanging data. The assessment results of the scheme approach those of the Tweedie regression model learned from centralized data, and outperform the Tweedie regression model learned independently by a single party.

<sec> <title>BACKGROUND</title> The health risks of phthalate esters are determined by the amount of exposure, individual sensitivities, and promoting factors. </sec> <sec> <title>OBJECTIVE</title> Using artificial intelligence algorithms and applied data mining to identify the correlations between phthalate esters [di(2-ethylhexyl) phthalate, DEHP], lifestyle, and disease </sec> <sec> <title>METHODS</title> This exploratory study collected basic demographic and laboratory data from the Taiwan Biobank. A prediction model for the relationship between phthalate esters and high disease risk was developed using several artificial intelligence algorithms, including logistic regression, an artificial neural network, and a Bayesian network. </sec> <sec> <title>RESULTS</title> The results revealed that phthalate esters have a greater influence on bone and joint problems, but a lesser impact on heart problems. On the other hand, the metabolites of DEHP, mono(2-carboxymethylhexyl) phthalate, mono-n-butyl phthalate, and monoethylphthalate, leave higher residues in females than in males, and the difference is statistically significant. Levels of monoethylphthalate are lower in people who exercise regularly than those who do not, and the difference is statistically significant. </sec> <sec> <title>CONCLUSIONS</title> The results of this study can provide a reference for the clinical nursing assessment of diseases related to osteoporosis, arthritis, and musculoskeletal pain, and facilitate medical staff to consider factors other than patients' basic physical assessment items, thereby improving medical care quality. </sec> <sec> <title>CLINICALTRIAL</title> NCT05892029, retrospectively registered on 19/05/2023. </sec>

This thesis presents the first measurement of prompt D+s mesons in heavy ion collisions with the CMS experiment. The transverse momentum (pT) spectra of prompt D+s mesons and charge conjugates are measured in pp and PbPb collisions at a center-of-mass energy of 5.02 TeV per nucleon pair using the CMS detector at the LHC. Themeasurement is performed through the D+s→φπ+→K+K−π+ decay channel with the D+s rapidity range |y| < 1.0. The D+s production in the pT range from 2 (6)GeV/c to 40 GeV/c in pp (PbPb) collisions is measured. Suppression of the D+s nuclear modification factor (RAA) in PbPb collisions suggests a significant interaction between charm quarks and the quark-gluon plasma. The double ratio of prompt D+s to prompt D0 production in pp and PbPb is measured. The ratio is consistent witha PHSD model calculation and consistent with unity, which indicates that strange charm meson enhancement in PbPb collisions is not found in the measured pT interval.

In our quest to understand the nature of dark matter and discover its non-gravitational interactions with ordinary matter, we propose an experiment using a \pbo ~calorimeter to search for or set new limits on the production rate of i) hidden sector particles in the $3 - 60$ MeV mass range via their $e^+e^-$ decay (or $\gamma\gamma$ decay with limited tracking), and ii) the hypothetical X17 particle, claimed in multiple recent experiments. The search for these particles is motivated by new hidden sector models and dark matter candidates introduced to account for a variety of experimental and observational puzzles: the small-scale structure puzzle in cosmological simulations, anomalies such as the 4.2$\sigma$ disagreement between experiments and the standard model prediction for the muon anomalous magnetic moment, and the excess of $e^+e^-$ pairs from the $^8$Be M1 and $^4$He nuclear transitions to their ground states observed by the ATOMKI group. In these models, the $1 - 100$ MeV mass range is particularly well-motivated and the lower part of this range still remains unexplored. Our proposed direct detection experiment will use a magnetic-spectrometer-free setup (the PRad apparatus) to detect all three final state particles in the visible decay of a hidden sector particle allowing for an effective control of the background and will cover the proposed mass range in a single setting. The use of the well-demonstrated PRad setup allows for an essentially ready-to-run and uniquely cost-effective search for hidden sector particles in the $3 - 60$ MeV mass range with a sensitivity of 8.9$\times$10$^{-8}$ - 5.8$\times$10$^{-9}$ to $\epsilon^2$, the square of the kinetic mixing interaction constant between hidden and visible sectors. This updated proposal includes our response to the PAC49 comments.

In a software-defined network (SDN), statistics information is of vital importance for different applications, such as traffic engineering, flow rerouting, and attack detection. Since some resources, e.g., ternary content addressable memory, SRAM, and computing capacity, are often limited in SDN switches, traffic measurements based on flow tables or sampling become infeasible. Sketch, a hash-based data structure, monitors every packet with fixed-size memory to provide a feasible approach of traffic measurement, but there exists a tradeoff between accuracy and memory. Currently, many efficient sketch algorithms have been designed to different purposes, but they focus on the performance and applications of one single sketch. In this paper, we present a scheme to reduce redundant flow statistics collected by sketches of different SDN switches. The proposed scheme could reduce measurement overhead in sketches, obtain more accurate estimate flow size, and find the elephant flow precisely.