An FM-Index Based High-Throughput Memory-Efficient FPGA Accelerator for Paired-End Short-Read Mapping.

This article presents an Ferragina-Manzini index (FM-index) based paired-end short-read mapping hardware accelerator. Four techniques are proposed to significantly reduce the number of memory accesses and operations to improve the throughput. First, an interleaved data structure is proposed to reduce the processing time by 51.8% by leveraging the data locality. Second, the boundaries of possible mapping location candidates can be retrieved within only one memory access by constructing a lookup table along with the FM-index. This reduces the number of DRAM accesses by 60% with only a 64 MB memory overhead. Third, an additional step is added to skip the time-consuming repetitive location candidates filtering conditionally, avoiding unnecessary operations. Lastly, an early termination method is proposed to terminate the mapping process if any location candidate with a high enough alignment score is detected, greatly decreasing the execution time. Overall, the computation time is reduced by 92.6% with only a 2% memory overhead in DRAM. The proposed methods are realized on a Xilinx Alveo U250 FPGA. The proposed FPGA accelerator processes 1,085,812,766 short-reads from the U.S. Food and Drug Administration (FDA) dataset within 35.4 minutes at 200 MHz. It achieves a 1.7-to-18.6× higher throughput and the highest 99.3% accuracy by exploiting the paired-end short-read mapping, compared to state-of-the-art FPGA-based designs.

Deep Unfolding for Topic Models.

Deep unfolding provides an approach to integrate the probabilistic generative models and the deterministic neural networks. Such an approach is benefited by deep representation, easy interpretation, flexible learning and stochastic modeling. This study develops the unsupervised and supervised learning of deep unfolded topic models for document representation and classification. Conventionally, the unsupervised and supervised topic models are inferred via the variational inference algorithm where the model parameters are estimated by maximizing the lower bound of logarithm of marginal likelihood using input documents without and with class labels, respectively. The representation capability or classification accuracy is constrained by the variational lower bound and the tied model parameters across inference procedure. This paper aims to relax these constraints by directly maximizing the end performance criterion and continuously untying the parameters in learning process via deep unfolding inference (DUI). The inference procedure is treated as the layer-wise learning in a deep neural network. The end performance is iteratively improved by using the estimated topic parameters according to the exponentiated updates. Deep learning of topic models is therefore implemented through a back-propagation procedure. Experimental results show the merits of DUI with increasing number of layers compared with variational inference in unsupervised as well as supervised topic models.

Color Doppler evaluation of blood flow changes in leiomyomas after uterine artery ligation.

OBJECTIVE: To detect possible changes in main blood vessels within leiomyomas after uterine artery ligation using color Doppler sonography. METHOD: Blood flow in main leiomyoma blood vessels was measured before and after the procedure in 14 women who also had abnormal uterine bleeding, pelvic pain or pressure, and/or anemia. RESULTS: Of the 14 patients, 13 reported complete disappearance of preoperative pain or pressure and 1 reported significant relief. Within 1 week to 4 months after uterine artery ligation, major blood flow within leiomyomas had significantly decreased in all patients. Eight months after the procedure, 1 of the women became pregnant. CONCLUSION: Laparoscopic uterine artery ligation via a lateral retroperitoneal technique is a safe and effective treatment for leiomyomas. Color Doppler sonography verified the ability of the procedure to diminish blood flow within leiomyomas in all patients.