RESUMO
Filter pruning has gained widespread adoption for the purpose of compressing and speeding up convolutional neural networks (CNNs). However, the existing approaches are still far from practical applications due to biased filter selection and heavy computation cost. This article introduces a new filter pruning method that selects filters in an interpretable, multiperspective, and lightweight manner. Specifically, we evaluate the contributions of filters from both individual and overall perspectives. For the amount of information contained in each filter, a new metric called information capacity is proposed. Inspired by the information theory, we utilize the interpretable entropy to measure the information capacity and develop a feature-guided approximation process. For correlations among filters, another metric called information independence is designed. Since the aforementioned metrics are evaluated in a simple but effective way, we can identify and prune the least important filters with less computation cost. We conduct comprehensive experiments on benchmark datasets employing various widely used CNN architectures to evaluate the performance of our method. For instance, on ILSVRC-2012, our method outperforms state-of-the-art methods by reducing floating-point operations (FLOPs) by 77.4% and parameters by 69.3% for ResNet-50 with only a minor decrease in an accuracy of 2.64%.
RESUMO
BACKGROUND: Some studies have found surgery and anesthesia in children to be associated with neurodevelopmental deficits, but specific reasons for this association have not been fully explored. This study evaluates intraoperative mean arterial pressure (MAP) during a single ambulatory procedure in children and subsequent mental disorder diagnoses. METHODS: A retrospective observational study was performed including children ≥28 days and <18 years of age with intraoperative electronic anesthetic records between January 1, 2009, and April 30, 2017, at our institution. Eligible children were categorized based on their mean intraoperative MAP relative to other children of the same sex and similar age: category 1 (very low): children with mean intraoperative MAP values below the 10th percentile, category 2 (low): mean MAP value ≥10th and <25th percentiles, category 3 (reference): mean MAP value ≥25th and <75th percentiles, category 4 (high): mean MAP value ≥75th and <90th percentile, and category 5 (very high): mean MAP value ≥90th percentile. International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9) and ICD, Tenth Revision, Clinical Modification (ICD-10)-coded mental disorders were identified in hospital and outpatient claims, with a median duration of follow-up after surgery of 120 days (interquartile range [IQR], 8-774.5 days). Cox proportional hazards models evaluated the hazard ratio (HR) of time to first mental disorder diagnosis associated with intraoperative blood pressure category between the end of surgery and censoring, with the primary analysis adjusting for demographic, anesthetic, comorbidity, and procedure-type variables as potential confounders. RESULTS: A total of 14,724 eligible children who received general anesthesia for a single ambulatory surgical procedure were identified. After adjusting for all available potential confounders, when compared to the reference, there were no statistically significant differences in mental disorder diagnosis risk based on intraoperative mean MAP category. Compared to reference, children in the very low and low blood pressure categories reported HRs of 1.00 (95% confidence interval [CI], 0.74-1.35) and 1.10 (95% CI, 0.87-1.41) for a mental disorder diagnosis, respectively, and children in the high and very high categories reported HRs of 0.87 (95% CI, 0.68-1.12) and 0.76 (95% CI, 0.57-1.03), respectively. CONCLUSIONS: Presence in a predefined mean intraoperative MAP category was not associated with subsequent mental disorder diagnoses within our follow-up period. However, the limitations of this study, including uncertainty regarding what constitutes an adequate blood pressure in children, may limit the ability to form definitive conclusions.