Search

We propose new versions of accelerated first-order methods for convex composite optimization, where the prox parameter is allowed to increase from one iteration to the next. In particular, we show that a full backtracking strategy can be used within the FISTA and FALM algorithms while preserving their worst-case iteration complexities of O([square root of L(f)/ζ]). In the original versions of FISTA and FALM the prox parameter value on each iteration must be bounded from above by its value on prior iterations. The complexity of the algorithm then depends on the smallest value of the prox parameter encountered by the algorithm. The theoretical implications of using full backtracking in the framework of accelerated first-order and alternating linearization methods allow for better complexity estimates that depend on the 'average' prox parameter value. Moreover, we show that in the case of compressed sensing problem and Lasso, the additional cost of the new backtracking strategy is negligible compared to the cost of the original FISTA iteration. Our computational results show the benefit of the new algorithm. Keywords Convex optimization * Composite optimization * Accelerated first-order methods * Convergence rates Mathematics Subject Classification 90C25, 1 Introduction First-order methods for convex optimization have become a focus of active research during the past several years. Their low per-iteration complexity and improved convergence rates have made them [...]

Hyperperfusion is related to vessel recanalization, tissue reperfusion, and collateral circulation. To determine the prognostic impact of hyperperfusion after an acute ischemic stroke (AIS) identified by arterial spin labeling (ASL) cerebral blood flow. Studies published in PubMed, Embase, and Cochrane Library databases were searched. Studies assessing the diagnostic performance of ASL hyperperfusion after AIS were included. Functional prognosis, hemorrhagic transformation (HT), infarction volume, and penumbra salvage volume were evaluated. The standardized mean difference or risk ratio was pooled, implementing a random effect model. Multiple subgroup analyses were performed. Seven studies including 617 participants were included in this meta-analysis. ASL hyperperfusion in AIS was correlated well with symptom severity and outcome after 24 h National Institutes of Health Stroke Scale (NIHSS) and 90-day modified Rankin Scale (mRS). Earlier ASL hyperperfusion was associated with a smaller infarction volume and a larger penumbra salvage volume, while also indicating a higher risk of HT. In addition, in subgroup analysis, our results demonstrated that thrombolysis, mechanical thrombectomy treatment, early improvement of NIHSS, and involving infarction in cortical territory are associated with ASL hyperperfusion. ASL hyperperfusion was related to a favorable functional outcome but an increased risk of HT. Stroke patients with hyperperfusion showed smaller infarction volume and larger penumbra salvage volume than those with non-hyperperfusion. [ABSTRACT FROM AUTHOR]

With the development of new methods and the tremendous progress in transducer technology, the observations and researches have become more and more stereoscopic and full-scale. In order to build the multi-source data fusion system propping up the computations, such as process evolution prediction, structure discovery and association analysis, the digital modelling of the natural entity needs to be carried out, which would help build the corresponding digital entity. In this study, the concepts and models in geoscience are introduced, and the issues overlooked in the digital modelling theories are discussed. On this basis, a unified conceptual model and its pseudo-representation (BPRModel) are built. Furthermore, the application of the model is illustrated under the research of specific natural entity, that is to say, the Earth's magnetosphere. [ABSTRACT FROM AUTHOR]