Hierarchical Pipeline Research Articles

Redundancy-Free and Load-Balanced TGNN Training With Hierarchical Pipeline Parallelism

Redundancy-Free and Load-Balanced TGNN Training With Hierarchical Pipeline Parallelism

Geochemistry π: Automated Machine Learning Python Framework for Tabular Data

AbstractAlthough machine learning (ML) has brought new insights into geochemistry research, its implementation is laborious and time‐consuming. Here, we announce Geochemistry π, an open‐source automated ML Python framework. Geochemists only need to provide tabulated data and select the desired options to clean data and run ML algorithms. The process operates in a question‐and‐answer format, and thus does not require that users have coding experience. After either automatic or manual parameter tuning, the automated Python framework provides users with performance and prediction results for the trained ML model. Based on the scikit‐learn library, Geochemistry π has established a customized automated process for implementing classification, regression, dimensionality reduction, and clustering algorithms. The Python framework enables extensibility and portability by constructing a hierarchical pipeline architecture that separates data transmission from the algorithm application. The AutoML module is constructed using the Cost‐Frugal Optimization and Blended Search Strategy hyperparameter search methods from the A Fast and Lightweight AutoML Library, and the model parameter optimization process is accelerated by the Ray distributed computing framework. The MLflow library is integrated into ML lifecycle management, which allows users to compare multiple trained models at different scales and manage the data and diagrams generated. In addition, the front‐end and back‐end frameworks are separated to build the web portal, which demonstrates the ML model and data science workflow through a user‐friendly web interface. In summary, Geochemistry π provides a Python framework for users and developers to accelerate their data mining efficiency with both online and offline operation options.

R2D2: Reducing Redundancy and Duplication in Data Lakes

Enterprise data lakes often suffer from substantial amounts of duplicate and redundant data, with data volumes ranging from terabytes to petabytes. This leads to both increased storage costs and unnecessarily high maintenance costs for these datasets. In this work, we focus on identifying and reducing redundancy in enterprise data lakes by addressing the problem of "dataset containment". To the best of our knowledge, this is one of the first works that addresses table-level containment at a large scale. We propose R2D2: a three-step hierarchical pipeline that efficiently identifies almost all instances of containment by progressively reducing the search space in the data lake. It first builds (i) a schema containment graph, followed by (ii) statistical min-max pruning, and finally, (iii) content level pruning. We further propose minimizing the total storage and access costs by optimally identifying redundant datasets that can be deleted (and reconstructed on demand) while respecting latency constraints. We implement our system on Azure Databricks clusters using Apache Spark for enterprise data stored in ADLS Gen2, and on AWS clusters for open-source data. In contrast to existing modified baselines that are inaccurate or take several days to run, our pipeline can process an enterprise customer data lake at the TB scale in approximately 5 hours with high accuracy. We present theoretical results as well as extensive empirical validation on both enterprise (scale of TBs) and open-source datasets (scale of MBs - GBs), which showcase the effectiveness of our pipeline.

Unsupervised Registration Refinement for Generating Unbiased Eye Atlas.

With the confounding effects of demographics across large-scale imaging surveys, substantial variation is demonstrated with the volumetric structure of orbit and eye anthropometry. Such variability increases the level of difficulty to localize the anatomical features of the eye organs for populational analysis. To adapt the variability of eye organs with stable registration transfer, we propose an unbiased eye atlas template followed by a hierarchical coarse-to-fine approach to provide generalized eye organ context across populations. Furthermore, we retrieved volumetric scans from 1842 healthy patients for generating an eye atlas template with minimal biases. Briefly, we select 20 subject scans and use an iterative approach to generate an initial unbiased template. We then perform metric-based registration to the remaining samples with the unbiased template and generate coarse registered outputs. The coarse registered outputs are further leveraged to train a deep probabilistic network, which aims to refine the organ deformation in unsupervised setting. Computed tomography (CT) scans of 100 de-identified subjects are used to generate and evaluate the unbiased atlas template with the hierarchical pipeline. The refined registration shows the stable transfer of the eye organs, which were well-localized in the high-resolution (0.5 mm3) atlas space and demonstrated a significant improvement of 2.37% Dice for inverse label transfer performance. The subject-wise qualitative representations with surface rendering successfully demonstrate the transfer details of the organ context and showed the applicability of generalizing the morphological variation across patients.

Hierarchical Point Cloud Encoding and Decoding With Lightweight Self-Attention Based Model

In this letter we present SA-CNN, a hierarchical and lightweight self-attention based encoding and decoding architecture for representation learning of point cloud data. The proposed SA-CNN introduces convolution and transposed convolution stacks to capture and generate contextual information among unordered 3D points. Following conventional hierarchical pipeline, the encoding process extracts feature in local-to-global manner, while the decoding process generates feature and point cloud in coarse-to-fine, multi-resolution stages. We demonstrate that SA-CNN is capable of a wide range of applications, namely classification, part segmentation, reconstruction, shape retrieval, and unsupervised classification. While achieving the state-of-the-art or comparable performance in the benchmarks, SA-CNN maintains its model complexity several order of magnitude lower than the others. In term of qualitative results, we visualize the multi-stage point cloud reconstructions and latent walks on rigid objects as well as deformable non-rigid human and robot models.

Safe, modular packet pipeline programming

The P4 language and programmable switch hardware, like the Intel Tofino, have made it possible for network engineers to write new programs that customize operation of computer networks, thereby improving performance, fault-tolerance, energy use, and security. Unfortunately, possible does not mean easy —there are many implicit constraints that programmers must obey if they wish their programs to compile to specialized networking hardware. In particular, all computations on the same switch must access data structures in a consistent order, or it will not be possible to lay that data out along the switch’s packet-processing pipeline. In this paper, we define Lucid 2.0, a new language and type system that guarantees programs access data in a consistent order and hence are pipeline-safe . Lucid 2.0 builds on top of the original Lucid language, which is also pipeline-safe, but lacks the features needed for modular construction of data structure libraries. Hence, Lucid 2.0 adds (1) polymorphism and ordering constraints for code reuse; (2) abstract, hierarchical pipeline locations and data types to support information hiding; (3) compile-time constructors, vectors and loops to allow for construction of flexible data structures; and (4) type inference to lessen the burden of program annotations. We develop the meta-theory of Lucid 2.0, prove soundness, and show how to encode constraint checking as an SMT problem. We demonstrate the utility of Lucid 2.0 by developing a suite of useful networking libraries and applications that exploit our new language features, including Bloom filters, sketches, cuckoo hash tables, distributed firewalls, DNS reflection defenses, network address translators (NATs) and a probabilistic traffic monitoring service.

A Reconfigurable Neural Network Processor With Tile-Grained Multicore Pipeline for Object Detection on FPGA

In order to improve the computational efficiency of convolutional neural networks (CNNs) for object detection on reconfigurable platforms such as field-programmable gate arrays (FPGAs), we propose a CNN processor with hierarchical pipelining and multicore reconfigurable computing based on parallel parameter constraints. First, we propose a pipelined multicore processing architecture that can adapt to computations and on-chip memory requirements of different convolutional layers. We present the design of a CNN processor that can configure the computing units while adjusting the interconnection of multicore to improve the utilization of reconfigurable computing resources. Second, we propose an elastic on-chip buffer and a data access approach by dynamically configuring addresses to better utilize on-chip memory. Meanwhile, we present a cross-layer feature map fusion strategy based on computing near memory (CNM) to reduce off-chip memory accesses. Finally, we propose a scheduling algorithm for pipelined tasks to improve the computational efficiency and throughput of the proposed processor. For evaluation, the well-known object detection methods (RetinaNet-ResNet-50, MobileNetV2-SSDLite, and YOLOv3) performed using the proposed CNN processor on the ZCU102 platform and reached the throughput of 1503, 1066, and 809 GOPS and the computational efficiency of 0.79, 0.62, and 0.36 GOPS/DSP, respectively. The designed processor realized a better tradeoff between computing efficiency and detection accuracy compared with the recently proposed object detection CNNs on FPGA.

Fast Implementation of Genetic Algorithm Based on Software/Hardware Co-design Method

Genetic algorithm (GA) has been widely used since it was proposed. However, due to the complexity of the algorithm itself, the fact that the actual problems have huge data and scale, and that it needs to meet certain real-time requirements, the acceleration of genetic algorithm has become a hot spot in both academia and industry field. Previous work have been done on the acceleration of GA, which primarily focuses on the software aspect, while the existing work in hardware aspect lacks adaptability for different industrial scenarios. Therefore, this paper puts forward a novel architecture based on software/hardware co-design method, in which the software produces random numbers, and the hardware, taking full account of the characteristics of sequential circuit, is implemented in a hierarchical pipeline architecture. This architecture accelerates the algorithm while providing certain flexibility. This paper completes the simulation on MATLAB and Vivado, and the total running time is 987.4.u. Compared with the pure software implementation, the speed is increased by 73.9 times.

Hierarchical Annotation Event Extraction Method in Multiple Scenarios

In the event extraction task, considering that there may be multiple scenarios in the corpus and an argument may play different roles under different triggers, the traditional tagging scheme can only tag each word once, which cannot solve the problem of argument overlap. A hierarchical tagging pipeline model for Chinese corpus based on the pretrained model Bert was proposed, which can obtain the relevant arguments of each event in a hierarchical way. The pipeline structure is selected in the model, and the event extraction task is divided into event trigger classification and argument recognition. Firstly, the pretrained model Bert is used to generate the feature vector and transfer it to bidirectional gated recurrent unit+conditional random field (BiGRU+CRF) model for trigger classification; then, the marked event type features are spliced into the corpus as known features and then passed into BiGRU+CRF for argument recognition. We evaluated our method on DUEE, combined with data enhancement and mask operation. Experimental results show that our method is improved compared with other baselines, which prove the effectiveness of the model in Chinese corpus.

Prognostic Stratification of Bladder Cancer Patients with a MicroRNA-based Approach.

Simple SummaryThe majority of patients with bladder cancer are diagnosed before the malignant cells invade the bladder’s muscle wall, and can be treated with surgery. These patients are nevertheless at high risk of disease recurrence and progression to muscle-invasive disease, and must undergo periodic cystoscopy and urine cytology, procedures that are burdensome for the patient and for the healthcare system. We set out to identify a follow-up/risk assessment test that analyzes the levels of a specific class of RNA molecules (microRNAs) in urine samples. Results led to the discovery of a panel of microRNAs in urine samples that identifies high-risk bladder cancer patients with high accuracy and predicts event-free survival. This urine microRNA assay thus holds promise as a noninvasive alternative to current methods for bladder cancer follow-up. Robust non-invasive tests for prognostic stratification of bladder cancer (BCa) patients are in high demand. Following a comprehensive analysis of studies on BCa, we selected a panel of 29 microRNAs (miRNAs) and analyzed their levels in urine and plasma samples in a prospective cohort of 63 BCa patients (32 at high risk of recurrence and 31 low-risk cases) and 37 healthy controls using RT-qPCR. To design an assay suitable for large-scale testing, we applied a hierarchical pipeline to select the miRNAs that were not affected by confounding factors such as haematuria and urine specific gravity, and exceeded stringent cut-off criteria (fold change > 2.5 and p-value < 0.005). Using a two-step decision tree based on the urine levels of miR-34a-5p, miR-200a-3p and miR-193a-5p, normalized against miR-125b-5p, patients could be classified as high- or low-risk with a sensitivity of 0.844, specificity of 0.806 and accuracy of 0.825. Furthermore, univariate Cox proportional hazards regression analyses indicated that increased urine levels of miR-29a-3p, miR-34a-5p, miR-193a-5p, miR-200c-3p, miR-205-5p and miR-532-5p were associated with a shorter event-free survival (hazard ratios > 3.1, p-value < 0.05). Taken together, our findings suggest that measuring the urine levels of these miRNAs could provide a novel cost-effective, noninvasive test for risk assessment of BCa patients.

Hierarchical visual localization for visually impaired people using multimodal images

Localization is one of the crucial issues in assistive technology for visually impaired people. In this paper, we propose a novel hierarchical visual localization pipeline based on the wearable assistive navigation device for visually impaired people. The proposed pipeline involves the deep descriptor network, 2D-3D geometric verification and online sequence matching. Images in different modalities (RGB, Infrared and Depth) are fed into Dual Desc network to generate robust attentive global descriptors and local features. The global descriptors are leveraged to retrieve the coarse candidates of query images. The 2D local features, as well as 3D sparse point cloud, are used in geometric verification to select the optimal results from the retrieved candidates. Finally, sequence matching robustifies the localization results by synthesizing the verified results of successive frames. The proposed unified descriptor network Dual Desc surpasses the state-of-the-art NetVLAD and its variant on the task of image description. Validated on the real-world dataset captured by the wearable assistive device, the proposed visual localization utilizes multimodal images to overcome the disadvantages of RGB images and robustifies the localization performance by deep descriptor network and hierarchical pipeline. In the challenging scenarios of the Yuquan dataset, the proposed method achieves the F1 score of 0.77 and the mean localization error of 2.75, which is satisfactory in practical use.

Design of drug-like hepsin inhibitors against prostate cancer and kidney stones

Hepsin, a transmembrane serine protease abundant in renal endothelial cells, is a promising therapeutic target against several cancers, particularly prostate cancer. It is involved in the release and polymerization of uromodulin in the urine, which plays a role in kidney stone formation. In this work, we design new potential hepsin inhibitors for high activity, improved specificity towards hepsin, and promising ADMET properties. The ligands were developed in silico through a novel hierarchical pipeline. This pipeline explicitly accounts for off-target binding to the related serine proteases matriptase and HGFA (human hepatocyte growth factor activator). We completed the pipeline incorporating ADMET properties of the candidate inhibitors into custom multi-objective optimization functions. The ligands designed show excellent prospects for targeting hepsin via the blood stream and the urine and thus enable key experimental studies. The computational pipeline proposed is remarkably cost-efficient and can be easily adapted for designing inhibitors against new drug targets.

Hierarchical Functional Maps between Subdivision Surfaces

AbstractWe propose a novel approach for computing correspondences between subdivision surfaces with different control polygons. Our main observation is that the multi‐resolution spectral basis functions that are open used for computing a functional correspondence can be compactly represented on subdivision surfaces, and therefore can be efficiently computed. Furthermore, the reconstruction of a pointwise map from a functional correspondence also greatly benefits from the subdivision structure. Leveraging these observations, we suggest a hierarchical pipeline for functional map inference, allowing us to compute correspondences between surfaces at fine subdivision levels, with hundreds of thousands of polygons, an order of magnitude faster than existing correspondence methods. We demonstrate the applicability of our results by transferring high‐resolution sculpting displacement maps and textures between subdivision models.

Efficient Interpretation of Deep Learning Models Using Graph Structure and Cooperative Game Theory: Application to ASD Biomarker Discovery.

Discovering imaging biomarkers for autism spectrum disorder (ASD) is critical to help explain ASD and predict or monitor treatment outcomes. Toward this end, deep learning classifiers have recently been used for identifying ASD from functional magnetic resonance imaging (fMRI) with higher accuracy than traditional learning strategies. However, a key challenge with deep learning models is understanding just what image features the network is using, which can in turn be used to define the biomarkers. Current methods extract biomarkers, i.e., important features, by looking at how the prediction changes if "ignoring" one feature at a time. However, this can lead to serious errors if the features are conditionally dependent. In this work, we go beyond looking at only individual features by using Shapley value explanation (SVE) from cooperative game theory. Cooperative game theory is advantageous here because it directly considers the interaction between features and can be applied to any machine learning method, making it a novel, more accurate way of determining instance-wise biomarker importance from deep learning models. A barrier to using SVE is its computational complexity: 2 N given N features. We explicitly reduce the complexity of SVE computation by two approaches based on the underlying graph structure of the input data: 1) only consider the centralized coalition of each feature; 2) a hierarchical pipeline which first clusters features into small communities, then applies SVE in each community. Monte Carlo approximation can be used for large permutation sets. We first validate our methods on the MNIST dataset and compare to human perception. Next, to insure plausibility of our biomarker results, we train a Random Forest (RF) to classify ASD/control subjects from fMRI and compare SVE results to standard RF-based feature importance. Finally, we show initial results on ranked fMRI biomarkers using SVE on a deep learning classifier for the ASD/control dataset.

Efficient Confidence-Based Hierarchical Stereo Disparity Upsampling for Noisy Inputs

Disparity upsampling techniques aim to restore high-resolution disparity maps from lowresolution disparity inputs. These inputs must be of high quality and are often obtained via complicated passive or active 3D reconstruction methods. Each pixel in the input disparity maps guides the disparity assignment in the upsampling process. The quality of the upsampled results will decrease if the initial disparity inputs are noisy, as the upsampled results are closely related to the initial inputs.We herein propose a hierarchical confidence-based upsampling framework that can be used to obtain relatively high quality upsampled results even under the noisy inputs. Specifically designed confidence measuring schemes are employed in our upsampling process, allowing the disparity assignment of only high-confidence pixels. For an effective depth quality evaluation, we present a novel classification of the confidence according to depth- and texture-related information and develop a confidence examination method with improved precision by combining multiple depth confidence evaluation methods. Our hierarchical pipeline contains 3 steps: confidence-based upsampling, confidence-based fine-tuning and confidence-based optimization. The upsampling combines multichannel information. Fine-tuning is carried out using the stereo texture information. Optimization is conducted utilizing the Markov random field method. All these proposed methods work together to suppress the low-confidence pixels and propagate the high-confidence pixels in the upsampling process. The cumulative error distribution is further analyzed, revealing the effectiveness of our confidence evaluation. Extensive comparison experiments are also performed using both the ground truth and stereo matching disparity maps as inputs to demonstrate the advantage of our framework over state-of-the-art upsampling methods.

A hierarchical pipeline for breast boundary segmentation and calcification detection in mammograms

Breast cancer is one of the most common cancer risks to women in the world. Amongst multiple breast imaging modalities, mammography has been widely used in breast cancer diagnosis and screening. Quantitative analyses including breast boundary segmentation and calcification localization are essential steps in a Computer Aided Diagnosis system based on mammography analysis. Due to uneven signal spatial distributions of pectoral muscle and glandular tissue, plus various artifacts in imaging, it is still challenging to automatically analyze mammogram images with high precision. In this paper, a fully automated pipeline of mammogram image processing is proposed, which estimates skin-air boundary using gradient weight map, detects pectoral-breast boundary by unsupervised pixel-wise labeling with no pre-labeled areas needed, and finally detects calcifications inside the breast region with a novel texture filter. Experimental results on Mammogram Image Analysis Society database show that the proposed method performs breast boundary segmentation and calcification detection with high accuracy of 97.08% and 96.15% respectively, and slightly higher accuracies are achieved on Full-Field Digital Mammography image datasets. Calculation of Jaccard and Dice indexes between segmented breast regions and the ground truths are also included as comprehensive similarity evaluations, which could provide valuable support for mammogram analysis in clinic.

Leveraging waveform complexity for confident detection of gravitational waves

The recent completion of Advanced LIGO suggests that gravitational waves (GWs) may soon be directly observed. Past searches for gravitational-wave transients have been impacted by transient noise artifacts, known as glitches, introduced into LIGO data due to instrumental and environmental effects. In this work, we explore how waveform complexity, instead of signal-to-noise ratio, can be used to rank event candidates and distinguish short duration astrophysical signals from glitches. We test this framework using a new hierarchical pipeline that directly compares the Bayesian evidence of explicit signal and glitch models. The hierarchical pipeline is shown to have strong performance, and in particular, allows high-confidence detections of a range of waveforms at realistic signal-to-noise ratio with a two detector network.

Segmentation of brain magnetic resonance images based on multi-atlas likelihood fusion: testing using data with a broad range of anatomical and photometric profiles.

We propose a hierarchical pipeline for skull-stripping and segmentation of anatomical structures of interest from T1-weighted images of the human brain. The pipeline is constructed based on a two-level Bayesian parameter estimation algorithm called multi-atlas likelihood fusion (MALF). In MALF, estimation of the parameter of interest is performed via maximum a posteriori estimation using the expectation-maximization (EM) algorithm. The likelihoods of multiple atlases are fused in the E-step while the optimal estimator, a single maximizer of the fused likelihoods, is then obtained in the M-step. There are two stages in the proposed pipeline; first the input T1-weighted image is automatically skull-stripped via a fast MALF, then internal brain structures of interest are automatically extracted using a regular MALF. We assess the performance of each of the two modules in the pipeline based on two sets of images with markedly different anatomical and photometric contrasts; 3T MPRAGE scans of pediatric subjects with developmental disorders vs. 1.5T SPGR scans of elderly subjects with dementia. Evaluation is performed quantitatively using the Dice overlap as well as qualitatively via visual inspections. As a result, we demonstrate subject-level differences in the performance of the proposed pipeline, which may be accounted for by age, diagnosis, or the imaging parameters (particularly the field strength). For the subcortical and ventricular structures of the two datasets, the hierarchical pipeline is capable of producing automated segmentations with Dice overlaps ranging from 0.8 to 0.964 when compared with the gold standard. Comparisons with other representative segmentation algorithms are presented, relative to which the proposed hierarchical pipeline demonstrates comparative or superior accuracy.

Hierarchical Pipelining of Nested Loops in High-Level Synthesis

Pipelining of the nested loops is very important in increasing the throughput of a system developed by a high-level synthesis tool. The most pipelining methods can handle only single loops. Therefore, nested loops are converted into a single loop, called loop flattened loop. In consequence, i.e. the sequential loops cannot be implemented in separate pipeline stages. This constraint limits the throughput. In this paper, a novel method are presented for nested loops by implementing to avoid this limitation. The method has the advantage that the desired restart time of the whole system can be given as an input parameter. The necessity of the pipeline scheduling on each loop hierarchy level can also be determined by this method. A novel multirate dataflow graph is also introduced for modeling the nested loops in an easy and abstract way.

Optimal Pipeline Paging Load Balancing for Hierarchical Cellular Networks

We study load balancing of paging schemes for multitier hierarchical cellular networks, in which different tiers of cells overlay each other to provide multiple coverage in cellular service areas. Each mobile terminal (MT) can be paged in any tier of a multitier hierarchical cellular network. Paging requests are balanced in different waiting queues of different tiers, and the load balancing among them is achieved probabilistically among N tiers. The studied paging schemes are the Hierarchical Pipeline Paging scheme, the Hierarchical Sequential Paging scheme, and the Hierarchical Blanket Paging scheme. We study two optimization problems using the N-tier load balancing: 1) given a paging delay constraint, to minimize the total paging cost under the constraint that the total delay is upper bounded by a predefined total delay, and 2) given a bound on the total delay, to minimize the total paging cost under a paging delay constraint.