scholarly journals Enhancing Urban Flow Maps via Neural ODEs

2020 ◽  
Author(s):  
Fan Zhou ◽  
Liang Li ◽  
Ting Zhong ◽  
Goce Trajcevski ◽  
Kunpeng Zhang ◽  
...  
Keyword(s):  
Spatial Correlation ◽  
Flow Distribution ◽  
Super Resolution ◽  
Traffic Monitoring ◽  
Coarse Grained ◽  
Flow Maps ◽  
Correlation Estimation ◽  
Gradient Computation ◽  
Memory Cost ◽  
Real World Datasets

Flow super-resolution (FSR) enables inferring fine-grained urban flows with coarse-grained observations and plays an important role in traffic monitoring and prediction. The existing FSR solutions rely on deep CNN models (e.g., ResNet) for learning spatial correlation, incurring excessive memory cost and numerous parameter updates. We propose to tackle the urban flows inference using dynamic systems paradigm and present a new method FODE -- FSR with Ordinary Differential Equations (ODEs). FODE extends neural ODEs by introducing an affine coupling layer to overcome the problem of numerically unstable gradient computation, which allows more accurate and efficient spatial correlation estimation, without extra memory cost. In addition, FODE provides a flexible balance between flow inference accuracy and computational efficiency. A FODE-based augmented normalization mechanism is further introduced to constrain the flow distribution with the influence of external factors. Experimental evaluations on two real-world datasets demonstrate that FODE significantly outperforms several baseline approaches.

Mechanisms of improvement in pulmonary gas exchange during isovolemic hemodilution

1999 ◽  
Vol 87 (1) ◽  
pp. 132-141 ◽  
Author(s):  
Steven Deem ◽  
Richard G. Hedges ◽  
Steven McKinney ◽  
Nayak L. Polissar ◽  
Michael K. Alberts ◽  
...  
Keyword(s):  
Blood Flow ◽  
Fractal Dimension ◽  
Gas Exchange ◽  
Spatial Correlation ◽  
Pulmonary Blood Flow ◽  
Minute Ventilation ◽  
Flow Distribution ◽  
Pulmonary Gas Exchange ◽  
Normal Lung ◽  
Blood Flow Distribution

Severe anemia is associated with remarkable stability of pulmonary gas exchange (S. Deem, M. K. Alberts, M. J. Bishop, A. Bidani, and E. R. Swenson. J. Appl. Physiol. 83: 240–246, 1997), although the factors that contribute to this stability have not been studied in detail. In the present study, 10 Flemish Giant rabbits were anesthetized, paralyzed, and mechanically ventilated at a fixed minute ventilation. Serial hemodilution was performed in five rabbits by simultaneous withdrawal of blood and infusion of an equal volume of 6% hetastarch; five rabbits were followed over a comparable time. Ventilation-perfusion (V˙a/Q˙) relationships were studied by using the multiple inert-gas-elimination technique, and pulmonary blood flow distribution was assessed by using fluorescent microspheres. Expired nitric oxide (NO) was measured by chemiluminescence. Hemodilution resulted in a linear fall in hematocrit over time, from 30 ± 1.6 to 11 ± 1%. Anemia was associated with an increase in arterial [Formula: see text] in comparison with controls ( P < 0.01 between groups). The improvement in O2 exchange was associated with reducedV˙a/Q˙heterogeneity, a reduction in the fractal dimension of pulmonary blood flow ( P = 0.04), and a relative increase in the spatial correlation of pulmonary blood flow ( P = 0.04). Expired NO increased with anemia, whereas it remained stable in control animals ( P < 0.0001 between groups). Anemia results in improved gas exchange in the normal lung as a result of an improvement in overallV˙a/Q˙matching. In turn, this may be a result of favorable changes in pulmonary blood flow distribution, as assessed by the fractal dimension and spatial correlation of blood flow and as a result of increased NO availability.

scholarly journals VAE-SNE: a deep generative model for simultaneous dimensionality reduction and clustering

2020 ◽  
Author(s):  
Jacob M. Graving ◽  
Iain D. Couzin
Keyword(s):  
Dimensionality Reduction ◽  
Image Data ◽  
Stereotyped Behavior ◽  
General Purpose ◽  
Coarse Grained ◽  
High Dimensional ◽  
Nonlinear Dimensionality Reduction ◽  
Multi Scale ◽  
Variational Autoencoder ◽  
Real World Datasets

AbstractScientific datasets are growing rapidly in scale and complexity. Consequently, the task of understanding these data to answer scientific questions increasingly requires the use of compression algorithms that reduce dimensionality by combining correlated features and cluster similar observations to summarize large datasets. Here we introduce a method for both dimension reduction and clustering called VAE-SNE (variational autoencoder stochastic neighbor embedding). Our model combines elements from deep learning, probabilistic inference, and manifold learning to produce interpretable compressed representations while also readily scaling to tens-of-millions of observations. Unlike existing methods, VAE-SNE simultaneously compresses high-dimensional data and automatically learns a distribution of clusters within the data — without the need to manually select the number of clusters. This naturally creates a multi-scale representation, which makes it straightforward to generate coarse-grained descriptions for large subsets of related observations and select specific regions of interest for further analysis. VAE-SNE can also quickly and easily embed new samples, detect outliers, and can be optimized with small batches of data, which makes it possible to compress datasets that are otherwise too large to fit into memory. We evaluate VAE-SNE as a general purpose method for dimensionality reduction by applying it to multiple real-world datasets and by comparing its performance with existing methods for dimensionality reduction. We find that VAE-SNE produces high-quality compressed representations with results that are on par with existing nonlinear dimensionality reduction algorithms. As a practical example, we demonstrate how the cluster distribution learned by VAE-SNE can be used for unsupervised action recognition to detect and classify repeated motifs of stereotyped behavior in high-dimensional timeseries data. Finally, we also introduce variants of VAE-SNE for embedding data in polar (spherical) coordinates and for embedding image data from raw pixels. VAE-SNE is a robust, feature-rich, and scalable method with broad applicability to a range of datasets in the life sciences and beyond.

scholarly journals Spectral Super-Resolution with Optimized Bands

2019 ◽  
Vol 11 (14) ◽  
pp. 1648 ◽  
Author(s):  
Utsav B. Gewali ◽  
Sildomar T. Monteiro ◽  
Eli Saber
Keyword(s):  
Network Architecture ◽  
Prior Information ◽  
Reflectance Spectrum ◽  
Super Resolution ◽  
High Dimensional ◽  
Sparse Signals ◽  
Reconstruction Accuracy ◽  
Neural Network Architecture ◽  
Real World Datasets ◽  
Very High

Hyperspectral (HS) sensors sample reflectance spectrum in very high resolution, which allows us to examine material properties in very fine details. However, their widespread adoption has been hindered because they are very expensive. Reflectance spectra of real materials are high dimensional but sparse signals. By utilizing prior information about the statistics of real HS spectra, many previous studies have reconstructed HS spectra from multispectral (MS) signals (which can be obtained from cheaper, lower spectral resolution sensors). However, most of these techniques assume that the MS bands are known apriori and do not optimize the MS bands to produce more accurate reconstructions. In this paper, we propose a new end-to-end fully convolutional residual neural network architecture that simultaneously learns both the MS bands and the transformation to reconstruct HS spectra from MS signals by analyzing large quantity of HS data. The learned band can be implemented in hardware to obtain an MS sensor that collects data that is best to reconstruct HS spectra using the learned transformation. Using a diverse set of real-world datasets, we show how the proposed approach of optimizing MS bands along with the transformation can drastically increase the reconstruction accuracy. Additionally, we also investigate the prospects of using reconstructed HS spectra for land cover classification.

Spatial correlation of regional pulmonary perfusion

1992 ◽  
Vol 72 (6) ◽  
pp. 2378-2386 ◽  
Author(s):  
R. W. Glenny
Keyword(s):  
Spatial Correlation ◽  
Three Dimensional ◽  
Flow Distribution ◽  
Pulmonary Perfusion ◽  
Heterogeneous Distribution ◽  
Mechanically Ventilated ◽  
Regional Flow ◽  
Flow Perfusion ◽  
Highly Correlated ◽  
Three Dimensional Coordinate

Despite the heterogeneous distribution of pulmonary blood flow, perfusion appears to be spatially ordered, with neighboring regions of lung having similar magnitudes of flow. This premise was tested by determining the spatial correlation of regional flow [rho(d)] as a function of distance (d) between regions. Regional pulmonary perfusion was measured in both supine and prone positions in seven anesthetized mechanically ventilated dogs with radiolabeled microspheres. After excision and drying, the lungs were cubed into pieces 1.2 cm on a side, with a three-dimensional coordinate assigned to each piece. The microsphere-determined flow to each piece was measured by radioactive counts, and rho(d) was calculated for all paired pieces within the same lobe. rho(d) was greatest for adjacent pieces (d = 1.2 cm) and decreased with increasing d, becoming negative at large distances in all dogs and positions. The spatial correlation of flow between adjacent pieces, rho(1.2 cm), was greater in the supine than in the prone position (0.66 vs. 0.72, P less than 0.05). The observations for each dog and position were fit to the equation rho(d) = d(a)+b.d+c, and the coefficients were used to compare rho(d) in the supine and prone positions. rho(d) differed in the two positions (P less than 0.05), with rho(d) falling off more rapidly with distance in the supine position. When trends in flow due to gravity were mathematically removed, differences between supine and prone positions were no longer observed. The spatial correlation of regional pulmonary perfusion was anisotropic in both supine and prone positions. The observation that regional pulmonary perfusion is highly correlated over large spatial distances has important implications for models of flow distribution.

scholarly journals Stratification relieves constraints from steric hindrance in the generation of compact actomyosin asters at the membrane cortex

2020 ◽  
Vol 6 (11) ◽  
pp. eaay6093
Author(s):  
Amit Das ◽  
Abrar Bhat ◽  
Rastko Sknepnek ◽  
Darius Köster ◽  
Satyajit Mayor ◽  
...  
Keyword(s):  
Myosin Ii ◽  
Super Resolution ◽  
Two Dimensions ◽  
Coarse Grained ◽  
In Vivo Studies ◽  
Agent Based ◽  
Molecular Stratification ◽  
Rectangular Slab

Recent in vivo studies reveal that several membrane proteins are driven to form nanoclusters by active contractile flows arising from localized dynamic patterning of F-actin and myosin at the cortex. Since myosin-II assemble as minifilaments with tens of myosin heads, one might worry that steric considerations would obstruct the emergence of nanoclustering. Using coarse-grained, agent-based simulations that account for steric constraints, we find that the patterns exhibited by actomyosin in two dimensions, do not resemble the steady-state patterns in our in vitro reconstitution of actomyosin on a supported bilayer. We perform simulations in a thin rectangular slab, separating the layer of actin filaments from myosin-II minifilaments. This recapitulates the observed features of in vitro patterning. Using super resolution microscopy, we find evidence for such stratification in our in vitro system. Our study suggests that molecular stratification may be an important organizing feature of the cortical cytoskeleton in vivo.

Sign in / Sign up

Export Citation Format

Share Document