Search

Geoscientific models are simplified representations of complex earth and environmental systems (EESs). Compared with physics-based numerical models, data-driven modeling has gained popularity due mainly to data proliferation in EESs and the ability to perform prediction without requiring explicit mathematical representation of complex biophysical processes. However, because of the black-box nature of data-driven models, their performance cannot be guaranteed. To address this issue, we developed a generalizable framework for improving the efficiency and effectiveness of model training and the reduction of model overfitting. This framework consists of two parts: hyperparameter selection based on Sobol global sensitivity analysis and hyperparameter tuning using a Bayesian optimization approach. We demonstrated the framework efficacy through a case study of daily edge-of-field (EOF) runoff predictions by a tree-based data-driven model using the extreme gradient boosting (XGBoost) algorithm in the Maumee domain, USA. This framework contributes towards improving the performance of a variety of data-driven models and can thus help promote their applications in EESs.

Geoscientific models are simplified representations of complex earth and environmental systems (EESs). Compared with physics-based numerical models, data-driven modeling has gained popularity due mainly to data proliferation in EESs and the ability to perform prediction without requiring explicit mathematical representation of complex biophysical processes. However, because of the black-box nature of data-driven models, their performance cannot be guaranteed. To address this issue, we developed a generalizable framework for the improvement of the efficiency and effectiveness of model training and the reduction of model overfitting. This framework consists of two parts: hyperparameter selection based on Sobol global sensitivity analysis, and hyperparameter tuning using a Bayesian optimization approach. We demonstrated the framework efficacy through a case study of daily edge-of-field (EOF) runoff predictions by a tree-based data-driven model using eXtreme Gradient Boosting (XGBoost) algorithm in the Maumee domain, U.S. This framework contributes towards improving the performance of a variety of data-driven models and can thus help promote their applications in EESs.

We showed in a phase I trial that the maximum tolerated dose of the ProMACE-CytaBOM regimen in patients with aggressive lymphoma was 200% (Gordon et al, J Clin Oncol 14:1275, 1996). Based on these observations, we initiated a phase II trial designed to determine response, toxicity, and dose intensity using this regimen. We analyzed 74 patients with advanced-stage (III or IV) or bulky stage II aggressive lymphoma. The overall complete response rate was 69% (72% in evaluable patients). With a median follow-up of 4.5 years, the median survival has not yet been reached. The 4-year survival rate is 73% (95% confidence interval [CI] 62, 83%) and no difference was observed among International Prognostic Index (IPI) groups. The 4-year disease-free survival was 71% (95% CI 58, 84%) with no statistical difference between patients with IPI 0 to 1 versus 2 to 4. The toxicity was acceptable, though the grade 4 hematologic toxicity rate for this regimen was 100%. Grade 4 nonhematologic toxicity was 36%. Three cases of either myelodysplastic syndrome or acute leukemia occurred at 7 months, 3.4 years, and 4.2 years after registration. Cytogenic analysis was available in two cases, showing inv(16) without French American British classification (FAB) M4 EO histology in one patient and a 5q-syndrome in the other. These data suggest that 200% ProMACE-CytaBOM with either granulocyte-macrophage colony-stimulating factor (GM-CSF) or G-CSF results in a high complete remission rate and a disease-free survival comparable to any prior risk-based analysis in aggressive lymphoma. Before using this regimen in general practice, phase III clinical trials should be conducted.

Background. This clinical trial was developed to determine whether ocular ice pack therapy would decrease 5-fluorouracil(5-FU)-induced ocular toxicity. Methods. Sixty-two patients who suffered from 5-FU-induced ocular toxicity, and were scheduled to receive another cycle of the chemotherapy that caused the ocular toxicity, were entered in this clinical trial. A randomized, crossover design was used, with patients documenting their ocular toxicity by the use of daily diaries. Results. The results from the first cycle of treatment suggested that ocular ice pack therapy decreased 5-FU-induced ocular toxicity ( P = 0.056). The 38 evaluable patients in the crossover analyses demonstrated decreased ocular toxicity with ocular ice pack therapy (p = .001). The ocular ice pack therapy was well tolerated by most of the study participants. Conclusion. Ocular ice pack therapy appears to lessen 5-FU-induced ocular toxicity to a clinically moderate degree. Better methods for decreasing 5-FU-induced ocular toxicity are necessary.

A 40-year-old woman presented with facial swelling and pressure sensation of the ears and sinuses. Chest roentgenography revealed a right paratracheal mass, which was confirmed by a venogram, and transjugular biopsy showed low-grade leiomyosarcoma. The superior vena cava was resected and reconstructed using a spiral vein graft. Pathologic evaluation revealed a low-grade leiomyosarcoma arising from both the superior vena cava and the azygos vein with clear margins.