Showing papers on "Sorting published in 2020"

Sparse Sinkhorn Attention

Abstract: We propose Sparse Sinkhorn Attention, a new efficient and sparse method for learning to attend. Our method is based on differentiable sorting of internal representations. Concretely, we introduce a meta sorting network that learns to generate latent permutations over sequences. Given sorted sequences, we are then able to compute quasi-global attention with only local windows, improving the memory efficiency of the attention module. To this end, we propose new algorithmic innovations such as Causal Sinkhorn Balancing and SortCut, a dynamic sequence truncation method for tailoring Sinkhorn Attention for encoding and/or decoding purposes. Via extensive experiments on algorithmic seq2seq sorting, language modeling, pixel-wise image generation, document classification and natural language inference, we demonstrate that our memory efficient Sinkhorn Attention method is competitive with vanilla attention and consistently outperforms recently proposed efficient Transformer models such as Sparse Transformers.

Raman image-activated cell sorting.

Abstract: The advent of image-activated cell sorting and imaging-based cell picking has advanced our knowledge and exploitation of biological systems in the last decade. Unfortunately, they generally rely on fluorescent labeling for cellular phenotyping, an indirect measure of the molecular landscape in the cell, which has critical limitations. Here we demonstrate Raman image-activated cell sorting by directly probing chemically specific intracellular molecular vibrations via ultrafast multicolor stimulated Raman scattering (SRS) microscopy for cellular phenotyping. Specifically, the technology enables real-time SRS-image-based sorting of single live cells with a throughput of up to ~100 events per second without the need for fluorescent labeling. To show the broad utility of the technology, we show its applicability to diverse cell types and sizes. The technology is highly versatile and holds promise for numerous applications that are previously difficult or undesirable with fluorescence-based technologies. Most current cell sorting methods are based on fluorescence detection with no imaging capability. Here the authors generate and use Raman image-activated cell sorting with a throughput of around 100 events per second, providing molecular images with no need for labeling.

Fast Differentiable Sorting and Ranking

Abstract: The sorting operation is one of the most commonly used building blocks in computer programming. In machine learning, it is often used for robust statistics. However, seen as a function, it is piecewise linear and as a result includes many kinks where it is non-differentiable. More problematic is the related ranking operator, often used for order statistics and ranking metrics. It is a piecewise constant function, meaning that its derivatives are null or undefined. While numerous works have proposed differentiable proxies to sorting and ranking, they do not achieve the $O(n \log n)$ time complexity one would expect from sorting and ranking operations. In this paper, we propose the first differentiable sorting and ranking operators with $O(n \log n)$ time and $O(n)$ space complexity. Our proposal in addition enjoys exact computation and differentiation. We achieve this feat by constructing differentiable operators as projections onto the permutahedron, the convex hull of permutations, and using a reduction to isotonic optimization. Empirically, we confirm that our approach is an order of magnitude faster than existing approaches and showcase two novel applications: differentiable Spearman's rank correlation coefficient and least trimmed squares.

Intelligent image-based deformation-assisted cell sorting with molecular specificity.

Abstract: Although label-free cell sorting is desirable for providing pristine cells for further analysis or use, current approaches lack molecular specificity and speed. Here, we combine real-time fluorescence and deformability cytometry with sorting based on standing surface acoustic waves and transfer molecular specificity to image-based sorting using an efficient deep neural network. In addition to general performance, we demonstrate the utility of this method by sorting neutrophils from whole blood without labels.

Mass Activated Droplet Sorting (MADS) Enables High-Throughput Screening of Enzymatic Reactions at Nanoliter Scale

Abstract: Microfluidic droplet sorting enables the high-throughput screening and selection of water-in-oil microreactors at speeds and volumes unparalleled by traditional well-plate approaches. Most such systems sort using fluorescent reporters on modified substrates or reactions that are rarely industrially relevant. We describe a microfluidic system for high-throughput sorting of nanoliter droplets based on direct detection using electrospray ionization mass spectrometry (ESI-MS). Droplets are split, one portion is analyzed by ESI-MS, and the second portion is sorted based on the MS result. Throughput of 0.7 samples s-1 is achieved with 98 % accuracy using a self-correcting and adaptive sorting algorithm. We use the system to screen ≈15 000 samples in 6 h and demonstrate its utility by sorting 25 nL droplets containing transaminase expressed in vitro. Label-free ESI-MS droplet screening expands the toolbox for droplet detection and recovery, improving the applicability of droplet sorting to protein engineering, drug discovery, and diagnostic workflows.

Intelligent image-activated cell sorting 2.0

Abstract: The advent of intelligent image-activated cell sorting (iIACS) has enabled high-throughput intelligent image-based sorting of single live cells from heterogeneous populations. iIACS is an on-chip microfluidic technology that builds on a seamless integration of a high-throughput fluorescence microscope, cell focuser, cell sorter, and deep neural network on a hybrid software-hardware data management architecture, thereby providing the combined merits of optical microscopy, fluorescence-activated cell sorting (FACS), and deep learning. Here we report an iIACS machine that far surpasses the state-of-the-art iIACS machine in system performance in order to expand the range of applications and discoveries enabled by the technology. Specifically, it provides a high throughput of ∼2000 events per second and a high sensitivity of ∼50 molecules of equivalent soluble fluorophores (MESFs), both of which are 20 times superior to those achieved in previous reports. This is made possible by employing (i) an image-sensor-based optomechanical flow imaging method known as virtual-freezing fluorescence imaging and (ii) a real-time intelligent image processor on an 8-PC server equipped with 8 multi-core CPUs and GPUs for intelligent decision-making, in order to significantly boost the imaging performance and computational power of the iIACS machine. We characterize the iIACS machine with fluorescent particles and various cell types and show that the performance of the iIACS machine is close to its achievable design specification. Equipped with the improved capabilities, this new generation of the iIACS technology holds promise for diverse applications in immunology, microbiology, stem cell biology, cancer biology, pathology, and synthetic biology.

An efficient image encryption using non-dominated sorting genetic algorithm-III based 4-D chaotic maps

Abstract: Chaotic maps are extensively utilized in the field of image encryption to generate secret keys. However, these maps suffer from hyper-parameters tuning issues. These parameters are generally selected on hit and trial basis. However, inappropriate selection of these parameters may reduce the performance of chaotic maps. Also, these hyper-parameters are not sensitive to input images. Therefore, in this paper, to handle these issues, a non-dominated sorting genetic algorithm-III (NSGA) based 4-D chaotic map is designed. Additionally, to improve the computational speed of the proposed approach, we have designed a novel master-slave model for image encryption. Initially, computationally expensive operations such as mutation and crossover of NSGA-III are identified. Thereafter, NSGA-III parameters are split among two jobs, i.e., master and slave jobs. For communication between master and slave nodes, the message passing interface is used. Extensive experimental results reveal that the proposed image encryption technique outperforms the existing techniques in terms of various performance measures.

Multidimensional Skills, Sorting, and Human Capital Accumulation ∗

Abstract: We construct a structural model of on-the-job search in which workers differ in skills along several dimensions (cognitive, manual, interpersonal...) and sort themselves into jobs with heterogeneous skill requirements along those same dimensions. We further allow for skills to be accumulated when used, and eroded away when not used. We estimate the model using occupation-level measures of skill requirements based on O*NET data, combined with a worker-level panel from the NLSY79. We use the estimated model to shed light on the origins and costs of mismatch along the cognitive, manual, and interpersonal skill dimensions. Our results clearly suggest that those three types of skills are very different productive attributes.

Image-Based Single Cell Sorting Automation in Droplet Microfluidics.

Abstract: The recent boom in single-cell omics has brought researchers one step closer to understanding the biological mechanisms associated with cell heterogeneity. Rare cells that have historically been obscured by bulk measurement techniques are being studied by single cell analysis and providing valuable insight into cell function. To support this progress, novel upstream capabilities are required for single cell preparation for analysis. Presented here is a droplet microfluidic, image-based single-cell sorting technique that is flexible and programmable. The automated system performs real-time dual-camera imaging (brightfield & fluorescent), processing, decision making and sorting verification. To demonstrate capabilities, the system was used to overcome the Poisson loading problem by sorting for droplets containing a single red blood cell with 85% purity. Furthermore, fluorescent imaging and machine learning was used to load single K562 cells amongst clusters based on their instantaneous size and circularity. The presented system aspires to replace manual cell handling techniques by translating expert knowledge into cell sorting automation via machine learning algorithms. This powerful technique finds application in the enrichment of single cells based on their micrographs for further downstream processing and analysis.

Emergency vehicle lane pre-clearing: From microscopic cooperation to routing decision making

Abstract: Emergency vehicles (EVs) play a crucial role in providing timely help for the general public in saving lives and avoiding property loss. However, very few efforts have been made for EV prioritization on normal road segments, such as the road section between intersections or highways between ramps. In this paper, we propose an EV lane pre-clearing strategy to prioritize EVs on such roads through cooperative driving with surrounding connected vehicles (CVs). The cooperative driving problem is formulated as a mixed-integer nonlinear programming (MINP) problem aiming at (i) guaranteeing the desired speed of EVs, and (ii) minimizing the disturbances on CVs. To tackle this NP-hard MINP problem, we formulate the model in a bi-level optimization manner to address these two objectives, respectively. In the lower-level problem, CVs in front of the emergency vehicle will be divided into several blocks. For each block, we developed an EV sorting algorithm to design optimal merging trajectories for CVs. With resultant sorting trajectories, a constrained optimization problem is solved in the upper-level to determine the initiation time/distance to conduct the sorting trajectories. Case studies show that with the proposed algorithm, emergency vehicles are able to drive at a desired speed while minimizing disturbances on normal traffic flows. We further reveal a linear relationship between the optimal solution and road density, which could help to improve EV routing decision makings when high-resolution data is not available.

Weakly-Supervised Crowd Counting Learns from Sorting Rather Than Locations

Abstract: In crowd counting datasets, the location labels are costly, yet, they are not taken into the evaluation metrics. Besides, existing multi-task approaches employ high-level tasks to improve counting accuracy. This research tendency increases the demand for more annotations. In this paper, we propose a weakly-supervised counting network, which directly regresses the crowd numbers without the location supervision. Moreover, we train the network to count by exploiting the relationship among the images. We propose a soft-label sorting network along with the counting network, which sorts the given images by their crowd numbers. The sorting network drives the shared backbone CNN model to obtain density-sensitive ability explicitly. Therefore, the proposed method improves the counting accuracy by utilizing the information hidden in crowd numbers, rather than learning from extra labels, such as locations and perspectives. We evaluate our proposed method on three crowd counting datasets, and the performance of our method plays favorably against the fully supervised state-of-the-art approaches.

Sequentially addressable dielectrophoretic array for high-throughput sorting of large-volume biological compartments.

Abstract: Droplet microfluidics has become a powerful tool in precision medicine, green biotechnology, and cell therapy for single-cell analysis and selection by virtue of its ability to effectively confine cells. However, there remains a fundamental trade-off between droplet volume and sorting throughput, limiting the advantages of droplet microfluidics to small droplets (<10 pl) that are incompatible with long-term maintenance and growth of most cells. We present a sequentially addressable dielectrophoretic array (SADA) sorter to overcome this problem. The SADA sorter uses an on-chip array of electrodes activated and deactivated in a sequence synchronized to the speed and position of a passing target droplet to deliver an accumulated dielectrophoretic force and gently pull it in the direction of sorting in a high-speed flow. We use it to demonstrate large-droplet sorting with ~20-fold higher throughputs than conventional techniques and apply it to long-term single-cell analysis of Saccharomyces cerevisiae based on their growth rate.

The Cross-Section of Risk and Returns

Abstract: A common practice in the finance literature is to create characteristic portfolios by sorting on characteristics associated with average returns. We show that the resultant portfolios are likely to capture not only the priced risk associated with the characteristic but also unpriced risk. We develop a procedure to remove this unpriced risk using covariance information estimated from past returns. We apply our methodology to the five Fama-French characteristic portfolios. The squared Sharpe ratio of the optimal combination of the resultant characteristic-efficient portfolios is 2.13, compared with 1.17 for the original characteristic portfolios.

Automatic image‐based identification and biomass estimation of invertebrates

Abstract: Understanding how biological communities respond to environmental changes is a key challenge in ecology and ecosystem management. The apparent decline of insect populations necessitates more biomonitoring but the time-consuming sorting and identification of taxa pose strong limitations on how many insect samples can be processed. In turn, this affects the scale of efforts to map invertebrate diversity altogether. Given recent advances in computer vision, we propose to replace the standard manual approach of human expert-based sorting and identification with an automatic image-based technology. We describe a robot-enabled image-based identification machine, which can automate the process of invertebrate identification, biomass estimation and sample sorting. We use the imaging device to generate a comprehensive image database of terrestrial arthropod species. We use this database to test the classification accuracy i.e. how well the species identity of a specimen can be predicted from images taken by the machine. We also test sensitivity of the classification accuracy to the camera settings (aperture and exposure time) in order to move forward with the best possible image quality. We use state-of-the-art Resnet-50 and InceptionV3 CNNs for the classification task. The results for the initial dataset are very promising ($\overline{ACC}=0.980$). The system is general and can easily be used for other groups of invertebrates as well. As such, our results pave the way for generating more data on spatial and temporal variation in invertebrate abundance, diversity and biomass.

Microfluidics for label-free sorting of rare circulating tumor cells.

Abstract: Circulating tumor cells (CTCs) have been widely considered as promising novel biomarkers for molecular research and clinical diagnosis of cancer. However, the sorting of CTCs is very challenging due to the rarity of CTCs in blood and the morphological similarity to blood cells. Although affinity-based CTC sorting methods could capture CTCs using specific biochemical markers, there are limitations such as the loss of cell viability after labeling and a requirement for expensive biochemical marker reagents. Emerging label-free CTC sorting methods rely on the physical properties of cells and can potentially overcome the aforementioned limitations. In this review, we highlight recent advances in label-free CTC sorting methods, with emphasis on device structures and performances. Specifically, we present a detailed discussion on label-free CTC sorting methods, including passive ones that depend on the channel structure or specific fluidic effects and active ones that use external force fields, as well as provide an overview of the principles, advantages, limitations, and applications of state-of-art label-free CTC sorting devices. Finally, we provide a future perspective of microfluidics for label-free CTC sorting and hope to inspire readers to develop new devices for applications in clinical cancer diagnoses and research.

FSPool: Learning Set Representations with Featurewise Sort Pooling

Abstract: Traditional set prediction models can struggle with simple datasets due to an issue we call the responsibility problem. We introduce a pooling method for sets of feature vectors based on sorting features across elements of the set. This can be used to construct a permutation-equivariant auto-encoder that avoids this responsibility problem. On a toy dataset of polygons and a set version of MNIST, we show that such an auto-encoder produces considerably better reconstructions and representations. Replacing the pooling function in existing set encoders with FSPool improves accuracy and convergence speed on a variety of datasets.

Spike detection and sorting with deep learning.

Abstract: Objective The extraction and identification of single-unit activities in intracortically recorded electric signals have a key role in basic neuroscience, but also in applied fields, like in the development of high-accuracy brain-computer interfaces. The purpose of this paper is to present our current results on the detection, classification and prediction of neural activities based on multichannel action potential recordings. Approach Throughout our investigations, a deep learning approach utilizing convolutional neural networks and a combination of recurrent and convolutional neural networks was applied, with the latter used in case of spike detection and the former used for cases of sorting and predicting spiking activities. Main results In our experience, the algorithms applied prove to be useful in accomplishing the tasks mentioned above: our detector could reach an average recall of 69%, while we achieved an average accuracy of 89% in classifying activities produced by more than 20 distinct neurons. Significance Our findings support the concept of creating real-time, high-accuracy action potential based BCIs in the future, providing a flexible and robust algorithmic background for further development.

Ensemble Many-Objective Optimization Algorithm Based on Voting Mechanism

Abstract: Sorting solutions play a key role in using evolutionary algorithms (EAs) to solve many-objective optimization problems (MaOPs). Generally, different solution-sorting methods possess different advantages in dealing with distinct MaOPs. Focusing on this characteristic, this article proposes a general voting-mechanism-based ensemble framework (VMEF), where different solution-sorting methods can be integrated and work cooperatively to select promising solutions in a more robust manner. In addition, a strategy is designed to calculate the contribution of each solution-sorting method and then the total votes are adaptively allocated to different solution-sorting methods according to their contribution. Solution-sorting methods that make more contribution to the optimization process are rewarded with more votes and the solution-sorting methods with poor contribution will be punished in a period of time, which offers a good feedback to the optimization process. Finally, to test the performance of VMEF, extensive experiments are conducted in which VMEF is compared with five state-of-the-art peer many-objective EAs, including NSGA-III, SPEA/R, hpaEA, BiGE, and grid-based evolutionary algorithm. Experimental results demonstrate that the overall performance of VMEF is significantly better than that of these comparative algorithms.

Combined Heat and Power Economic Emission Dispatch Using Hybrid NSGA II-MOPSO Algorithm Incorporating an Effective Constraint Handling Mechanism

Abstract: This research work proposes a synergistic hybrid metaheuristic algorithm a merger of Nondominated Sorting Genetic Algorithm II and Multiobjective Particle Swarm Optimization algorithm for solving the highly complicated combined heat and power economic emission dispatch problem to operate the power system economically and to reduce the impact of environmental pollution. During the iteration, based on ranking, the population is divided into two halves. The exploration is carried out by Nondominated Sorting Genetic Algorithm II using the upper half of the population. The modification of Multiobjective Particle Swarm Optimization to effectively exploit the lower half of the population is done by increasing the personal learning coefficient, decreasing the global learning coefficient and by using an adaptive mutation operator. To satisfy the linear, nonlinear constraints, and to ensure the populations always lie in the Feasible Operating Region of the cogeneration plant, an effective constraint handling mechanism is developed. The proposed hybrid algorithm with an effective constraint handling mechanism enhances the searching capability by effective information interchange. The algorithm is applied to standard test functions and test systems while considering the valve point effects of the thermal plants, transmission power losses, bounds of the units and feasible operating region of the cogeneration units. The hybrid algorithm can obtain a well spread and diverse Pareto optimal solution and also can converge to the actual Pareto optimal front faster than some of the existing algorithms. The statistical analysis reveals that the proposed hybrid algorithm is a viable alternative to solve this complicated and vital problem.

Analytic hierarchy process-fuzzy sorting: An analytic hierarchy process–based method for fuzzy classification in sorting problems

Abstract: Analytic Hierarchy Process (AHP) is a well-founded and popular method in the Multi-Criteria Decision Analysis (MCDA) field. AHPSort, a recently introduced sorting variant, uses crisp class-assignme...

Evaluation of image processing technique and discriminant analysis methods in postharvest processing of carrot fruit

Abstract: The most important process before packaging and preserving agricultural products is sorting operation. Sort of carrot by human labor is involved in many problems such as high cost and product waste. Image processing is a modern method, which has different applications in agriculture including classification and sorting. The aim of this study was to classify carrot based on shape using image processing technique. For this, 135 samples with different regular and irregular shapes were selected. After image acquisition and preprocessing, some features such as length, width, breadth, perimeter, elongation, compactness, roundness, area, eccentricity, centroid, centroid nonhomogeneity, and width nonhomogeneity were extracted. After feature selection, linear discriminant analysis (LDA) and quadratic discriminant analysis (QDA) methods were used to classify the features. The classification accuracies of the methods were 92.59 and 96.30, respectively. It can be stated that image processing is an effective way in improving the traditional carrot sorting techniques.

MorphoCluster: Efficient Annotation of Plankton Images by Clustering.

Abstract: In this work, we present MorphoCluster, a software tool for data-driven, fast, and accurate annotation of large image data sets. While already having surpassed the annotation rate of human experts, volume and complexity of marine data will continue to increase in the coming years. Still, this data requires interpretation. MorphoCluster augments the human ability to discover patterns and perform object classification in large amounts of data by embedding unsupervised clustering in an interactive process. By aggregating similar images into clusters, our novel approach to image annotation increases consistency, multiplies the throughput of an annotator, and allows experts to adapt the granularity of their sorting scheme to the structure in the data. By sorting a set of 1.2 M objects into 280 data-driven classes in 71 h (16 k objects per hour), with 90% of these classes having a precision of 0.889 or higher. This shows that MorphoCluster is at the same time fast, accurate, and consistent; provides a fine-grained and data-driven classification; and enables novelty detection.

Online Sorting of the Film on Cotton Based on Deep Learning and Hyperspectral Imaging

Abstract: Mulch film is usually mixed in with cotton during machine-harvesting and processing, which reduces the cotton quality. This paper presents a novel sorting algorithm for the online detection of film on cotton using hyperspectral imaging with a spectral region of 1000 - 2500 nm. The sorting algorithm consists of a group of stacked autoencoders, two optimization modules and an extreme learning machine (ELM) classifier. The variable-weighted stacked autoencoders (VW-SAE) are constructed to extract the features from hyperspectral images, and an artificial neural network (ANN), which is one optimization module, is applied to optimize the parameters of the VW-SAE. Then, the extracted features are input in the ELM to classify four types of objects: background, film on background, cotton and film on cotton. The ELM is optimized by a new optimizer (grey wolf optimizer), which can adjust the hidden nodes and parameters of the ELM simultaneously. A group of experiments was carried out to evaluate the performance of the proposed sorting algorithm using cotton that was provided by a Xinjiang municipality cotton ginning company. The experimental results show that the VW-SAE can improve the classification accuracies by approximately 15%. The overall recognition rate of the proposed algorithm is over 95%, and its recognition time is comparable to some state-of-the-art methods.

Dynamically tunable elasto-inertial particle focusing and sorting in microfluidics

Abstract: Inertial particle separation using passive hydrodynamic forces has attracted great attention in the microfluidics community because of its operation simplicity and high throughput sample processing. Due to the passive nature of inertial microfluidics, each inertial sorting device is typically fixed to a certain cut-off size for particle separation that is mainly dependent on the channel geometry and dimensions, which however lacks tunability in the separation threshold to fulfill the needs of different sorting applications. In this work, we explore the use of non-Newtonian viscoelastic fluids to achieve size-tunable elasto-inertial particle focusing and sorting in a microfluidic device with reverse wavy channel structures. The balance and competition among inertial lift force, Dean drag force and the controllable elastic lift force give rise to interesting size-based particle focusing phenomena with tunability in the equilibrium focusing positions. Seven differently sized fluorescent microspheres (0.3, 2, 3, 5, 7, 10 and 15 μm) are used to investigate the effects of the flow rate, viscoelastic fluid concentration and particle size on the tunable elasto-inertial focusing behavior. With the sorting tunability, we have achieved a highly effective sorting of a particle mixture into three subpopulations based on the particle size, i.e., small, intermediate and large subpopulations. We even demonstrate the controllable tunability among three separation thresholds for elasto-inertial particle sorting without changing the geometry and dimensions of the microfluidic device. The tunability of the developed elasto-inertial particle focusing and sorting can significantly broaden its application in a variety of biomedical research studies.

Optimal Spatial Policies, Geography, and Sorting

Abstract: We study optimal spatial policies in quantitative trade and geography frameworks with spillovers and sorting of heterogeneous workers. We rst characterize ecient spatial transfers and the labor subsidies that would implement them. Then, we quantify the aggregate and distributional eects of implementing these policies in the U.S. economy. Under homogeneous workers and constant-elasticity spillovers, a constant labor subsidy over space restores efficiency regardless of micro heterogeneity in fundamentals and trade costs. In that case, the quantification suggests that the observed spatial transfers in the U.S. are close to ecient. Spillovers across heterogeneous workers create an additional rationale for place-specific subsidies to attain optimal sorting. Under heterogeneous workers, the quantication suggests that optimal spatial policies may require stronger redistribution towards low-wage cities than in the data, reduce wage inequality in larger cities, weaken spatial sorting by skill, and lead to signicant welfare gains. Spillovers across dierent types of workers are a key driving force behind these results.