TimeTuner: Diagnosing Time Representations for Time-Series Forecasting with Counterfactual Explanations.

Deep learning (DL) approaches are being increasingly used for time-series forecasting, with many efforts devoted to designing complex DL models. Recent studies have shown that the DL success is often attributed to effective data representations, fostering the fields of feature engineering and representation learning. However, automated approaches for feature learning are typically limited with respect to incorporating prior knowledge, identifying interactions among variables, and choosing evaluation metrics to ensure that the models are reliable. To improve on these limitations, this paper contributes a novel visual analytics framework, namely TimeTuner, designed to help analysts understand how model behaviors are associated with localized correlations, stationarity, and granularity of time-series representations. The system mainly consists of the following two-stage technique: We first leverage counterfactual explanations to connect the relationships among time-series representations, multivariate features and model predictions. Next, we design multiple coordinated views including a partition-based correlation matrix and juxtaposed bivariate stripes, and provide a set of interactions that allow users to step into the transformation selection process, navigate through the feature space, and reason the model performance. We instantiate TimeTuner with two transformation methods of smoothing and sampling, and demonstrate its applicability on real-world time-series forecasting of univariate sunspots and multivariate air pollutants. Feedback from domain experts indicates that our system can help characterize time-series representations and guide the feature engineering processes.

"Parenchyma transection-first" strategy is superior to "tunnel-first" strategy in robotic spleen-preserving distal pancreatectomy with conservation of splenic vessels.

BACKGROUND: Creating a tunnel between the pancreas and splenic vessels followed by pancreatic parenchyma transection ("tunnel-first" strategy) has long been used in spleen-preserving distal pancreatectomy (SPDP) with splenic vessel preservation (Kimura's procedure). However, the operation space is limited in the tunnel, leading to the risks of bleeding and difficulties in suturing. We adopted the pancreatic "parenchyma transection-first" strategy to optimize Kimura's procedure. METHODS: The clinical data of consecutive patients who underwent robotic SPDP with Kimura's procedure between January 2017 and September 2022 at our center were retrieved. The cohort was classified into a "parenchyma transection-first" strategy (P-F) group and a "tunnel-first" strategy (T-F) group and analyzed. RESULTS: A total of 91 patients were enrolled in this cohort, with 49 in the T-F group and 42 in the P-F group. Compared with the T-F group, the P-F group had significantly shorter operative time (146.1 ± 39.2 min vs. 174.9 ± 46.6 min, P < 0.01) and lower estimated blood loss [40.0 (20.0-55.0) mL vs. 50.0 (20.0-100.0) mL, P = 0.03]. Failure of splenic vessel preservation occurred in 10.2% patients in the T-F group and 2.4% in the P-F group (P = 0.14). The grade 3/4 complications were similar between the two groups (P = 0.57). No differences in postoperative pancreatic fistula, abdominal infection or hemorrhage were observed between the two groups. CONCLUSIONS: The pancreatic "parenchyma transection-first" strategy is safe and feasible compared with traditional "tunnel-first strategy" in SPDP with Kimura's procedure.

Effects of service attributes and competition on electronic word of mouth: an elaboration likelihood perspective.

The management of electronic word of mouth (eWOM) is critical in e-commerce. In this study, on the basis of the elaboration likelihood model (ELM), we constructed a model of factors influencing eWOM by dividing merchants' attributes into the central and peripheral routes, which correspond to consumers' systematic and heuristic cognitive modes respectively. We then tested the developed model by using a cross-sectional data set. The results of this study indicate that the degree of competition faced by merchants has a significant negative association with eWOM. Moreover, price level and location moderate the relationship between competition and eWOM. The services of reservation and group buying have positive associations with eWOM. This research has three main contributions. First, we explored the effect of competition on eWOM. Second, we validated the feasibility of applying the ELM to the catering industry by dividing merchant attributes into the central and peripheral routes; this approach is consistent with systematic and heuristic cognitive theories. Finally, this research provides practical suggestions for eWOM management in the catering industry.

Evolution of laser-induced bulk damage in KDP crystal.

Pinpoint damage is the main type of bulk damage in potassium dihydrogen phosphate (KDP) crystals in high-power lasers. Using time-resolved microimaging, we observed the complete dynamic evolution of pinpoint damage in a KDP crystal. We analyzed changes in the patterns of dark zones formed by decreasing probe transmittance in transient images throughout the process. The mechanical properties of stress waves in KDP crystals were further studied by a depolarized shadowgraph experiment and theoretical simulation. The dynamic evolution of mechanical stress waves was observed, and the correlation between mechanical failure due to stress waves and the static characteristic damage morphology was established.

Functional Subdivisions of the Cerebellum in Naturalistic Paradigm Functional Magnetic Resonance Imaging.

Compelling evidence has suggested that the human cerebellum is engaged in a wide range of cognitive tasks besides traditional opinions of motor control, and it is organized into a set of distinct functional subregions. The existing model-driven cerebellum parcellations through resting-state functional MRI (rsfMRI) and task-fMRI are relatively coarse, introducing challenges in resolving the functions of the cerebellum especially when the brain is exposed to naturalistic environments. The current study took the advantages of the naturalistic paradigm (i.e., movie viewing) fMRI (nfMRI) to derive fine parcellations via a data-driven dual-regression-like sparse representation framework. The parcellations were quantitatively evaluated by functional homogeneity, and global and local boundary confidence. In addition, the differences of cerebellum-cerebrum functional connectivities between rsfMRI and nfMRI for some exemplar parcellations were compared to provide qualitatively functional validations. Our experimental results demonstrated that the proposed study successfully identified distinct subregions of the cerebellum. This fine parcellation may serve as a complementary solution to existing cerebellum parcellations, providing an alternative template for exploring neural activities of the cerebellum in naturalistic environments.

The development of Hani peatland in the Changbai mountains (NE China) and its response to the variations of the East Asian summer monsoon.

The past ecosystem responses to climate variability makes it possible to view the sensitivity of ecosystems to climate-forced state shifts. To test the hypothesis that the development of peatland in the Changbai Mountains responds to the variability of the East Asian summer monsoon (EASM), the developmental history of the Hani peatland was investigated based on peat basal ages. It can be concluded that the development of Hani peatland is the paludification. The development of the northern region started 13,685 cal. yr BP, while that of the southern region was initiated 7705 cal. yr BP. In addition, the moisture changes and development of the Hani peatland can be divided into three periods. From 16 to 7 ka cal. BP, the increase in the EASM was induced by the lower sea level and southward displacement of the Western Pacific Subtropical High (WPSH), resulting in increasing moisture in the Hani peatland. Peat was gradually established in the northern region. From 7 to 2 ka cal. BP, the increased relative sea level and northward displacement of the WPSH induced the maximum EASM. The EASM and Northeast Monsoon (NEM) induced the maximum moisture in the Hani peatland. Both the northern and southern regions were covered with peat during this period. Since 2 ka cal. BP, the decreasing EASM might be related to the seasonal decrease in the Northern Hemisphere summer insolation and ENSO intensity. Thus, the decreasing EASM induced the decrease in the moisture in the Hani peatland. Peat further accumulated in both the northern and southern regions. This study will help to understand the future EASM behavior in NE China and the development of similar peatlands in response to ongoing and future climatic change.