Gaps often occur in eddy covariance flux measurements, leading to data loss and necessitating accurate gap-filling. Furthermore, gaps in evapotranspiration (ET) measurements of annual field crops are ...particularly challenging to fill because crops undergo rapid change over a short season. In this study, an innovative deep learning (DL) gap-filling method was tested on a database comprising six datasets from different crops (cotton, tomato, and wheat). For various gap scenarios, the performance of the method was compared with the common gap-filling technique, marginal distribution sampling (MDS), which is based on lookup tables. Furthermore, a predictor importance analysis was performed to evaluate the importance of the different meteorological inputs in estimating ET. On the half-hourly time scale, the deep learning method showed a significant 13.5% decrease in nRMSE (normalized root mean square error) throughout all datasets and gap durations. A substantially smaller standard deviation of mean nRMSE, compared with marginal distribution sampling, was also observed. On the whole-gap time scale (half a day to six days), average nMBE (normalized mean bias error) was similar to that of MDS, whereas standard deviation was improved. Using only air temperature and relative humidity as input variables provided an RMSE that was significantly smaller than that resulting from the MDS method. These results suggest that the deep learning method developed here is reliable and more consistent than the standard gap-filling method and thereby demonstrates the potential of advanced deep learning techniques for improving dynamic time series modeling.
This paper revisits system identification and shows how new paradigms from machine learning can be used to improve it in the case of non-linear systems modeling from noisy and unbalanced dataset. We ...show that using importance sampling schemes in system identification can provide a significant performance boost in modeling, which is helpful to a predictive controller. The performance of the approach is first evaluated on simulated data of a Unmanned Surface Vehicle (USV). Our approach consistently outperforms baseline approaches on this dataset. Moreover we demonstrate the benefits of this identification methodology in a control setting. We use the model of the Unmanned Surface Vehicle (USV) in a Model Predictive Path Integral (MPPI) controller to perform a track following task. We discuss the influence of the controller parameters and show that the prioritized model outperform standard methods. Finally, we apply the Model Predictive Path Integral (MPPI) on a real system using the know-how developed here.
According to the industry, the value of wood logs is heavily influenced by their internal structure, particularly the distribution of knots within the trees. Nowadays, CT scanners combined with ...classical computer vision approach are the most common tool for obtaining reliable and accurate images of the interior structure of trees. Knowing where the tree semantic features, especially knots, contours and centers are within a tree could improve the efficiency of the overall tree industry by minimizing waste and enhancing the quality of wood-log by-products. However, this requires to automatically process the CT-scanner images so as to extract the different elements such as tree centerline, knot localization and log contour, in a robust and efficient manner. In this paper, we propose an effective methodology based on deep learning for performing these different tasks by processing CT-scanner images with deep convolutional neural networks. To meet this objective, three end-to-end trainable pipelines are proposed. The first pipeline is focused on centers detection using CNNs architecture with a regression head, the second and the third one address contour estimation and knot detection as a binary segmentation task based on an Encoder-Decoder architecture. The different architectures are tested on several tree species. With these experiments, we demonstrate that our approaches can be used to extract the different elements of trees in a precise manner while preserving good performances of robustness. The main objective was to demonstrate that methods based on deep learning might be used and have a relevant potential for segmentation and regression on CT-scans of tree trunks.
•DL pipeline was developed to detect semantic features in tree logs from Xray CT scan.•These approaches aim to automate the detection while preserving performances.•Demonstration of robustness and efficiency of these approaches on tree centerline.•Demonstration of robustness and efficiency of these approaches on knots, contours.
How to Train Your HERON Richard, Antoine; Aravecchia, Stephanie; Schillaci, Thomas ...
IEEE robotics and automation letters,
07/2021, Letnik:
6, Številka:
3
Journal Article
Recenzirano
In this letter we apply Deep Reinforcement Learning (Deep RL) and Domain Randomization to solve a navigation task in a natural environment relying solely on a 2D laser scanner. We train a model-based ...RL agent in simulation to follow lake and river shores and apply it on a real Unmanned Surface Vehicle in a zero-shot setup. We demonstrate that even though the agent has not been trained in the real world, it can fulfill its task successfully and adapt to changes in the robot's environment and dynamics. Finally, we show that the RL agent is more robust, faster, and more accurate than a state-aware Model-Predictive-Controller. Code, simulation environments, pre-trained models, and datasets are available at https://github.com/AntoineRichard/Heron-RL-ICRA.git .
While nonanadromous males (stream‐resident and/or mature male parr) contribute to reproduction in anadromous salmonids, little is known about their impacts on key population genetic parameters. Here, ...we evaluated the contribution of Atlantic salmon mature male parr to the effective number of breeders (Nb) using both demographic (variance in reproductive success) and genetic (linkage disequilibrium) methods, the number of alleles, and the relatedness among breeders. We used a recently published pedigree reconstruction of a wild anadromous Atlantic salmon population in which 2548 fry born in 2010 were assigned parentage to 144 anadromous female and 101 anadromous females that returned to the river to spawn in 2009 and to 462 mature male parr. Demographic and genetic methods revealed that mature male parr increased population Nb by 1.79 and 1.85 times, respectively. Moreover, mature male parr boosted the number of alleles found among progenies. Finally, mature male parr were in average less related to anadromous females than were anadromous males, likely because of asynchronous sexual maturation between mature male parr and anadromous fish of a given cohort. By increasing Nb and allelic richness, and by decreasing inbreeding, the reproductive contribution of mature male parr has important evolutionary and conservation implications for declining Atlantic salmon populations.
•BERT (Bidirectional Encoder Representations from Transformers)-based NLP (Natural Language Processing) model allows obtaining good performances with F1-score > 98% on French thoracic CT ...reports.•Accuracy of the NLP for identification of pulmonary embolism on CT reports is 99.1%.•15.8% of the CTs requested for suspected pulmonary embolism were positive.
To develop a Natural Language Processing (NLP) method based on Bidirectional Encoder Representations from Transformers (BERT) adapted to French CT reports and to evaluate its performance to calculate the diagnostic yield of CT in patients with clinical suspicion of pulmonary embolism (PE).
All the CT reports performed in our institution in 2019 (99,510 reports, training and validation dataset) and 2018 (94,559 reports, testing dataset) were included after anonymization. Two BERT-based NLP sentence classifiers were trained on 27.700, manually labeled, sentences from the training dataset. The first one aimed to classify the reports’ sentences into three classes (“Non chest”, “Healthy chest”, and "Pathological chest" related sentences), the second one to classify the last class into eleven sub classes pathologies including "pulmonary embolism". F1-score was reported on the validation dataset. These NLP classifiers were then applied to requested CT reports for pulmonary embolism from the testing dataset. Sensitivity, specificity, and accuracy for detection of the presence of a pulmonary embolism were reported in comparison to human analysis of the reports.
The F1-score for the 3-Classes and 11-SubClasses classifiers was 0.984 and 0.985, respectively. 4,042 examinations from the testing dataset were requested for pulmonary embolism of which 641 (15.8%) were positively evaluated by radiologists. The sensitivity, specificity, and accuracy of the NLP network for identifying pulmonary embolism in these reports were 98.2%, 99.3% and 99.1%, respectively.
BERT-based NLP sentences classifier enables the analysis of large databases of radiological reports to accurately determine the diagnostic yield of CT screening.
The incorporation of lipophilic ligands into the bilayer membrane of vesicles offers the possibility to induce, upon binding of suitable metal ions, a variety of processes, in particular vesicle ...aggregation and fusion and generation of vesicle arrays, under the control of specific metal-ligand recognition events. Synthetic bipyridine lipoligands Bn bearing a bipyridine unit as head group were prepared and incorporated into large unilamellar vesicles. The addition of Ni2+ or Co2+ metal ions led to the formation of complexes MBn and MBn2 followed by spontaneous fusion to generate giant multilamellar vesicles. The metal ion complexation was followed by UV spectroscopy and the progressive fusion could be visualized by optical dark-field and fluorescence microscopies. Vesicle fusion occurred without leakage of the aqueous compartments and resulted in the formation of multilamellar giant vesicles because of the stacking of the lipoligands Bn. The fusion process required a long enough oligoethylene glycol spacer and a minimal concentration of lipoligand within the vesicle membrane. Metallosupramolecular systems such as the present one offer an attractive way to induce selective intervesicular processes, such as vesicle fusion, under the control of molecular recognition between specific metal ions and lipoligands incorporated in the bilayer membrane. They provide an approach to the design of artificial "tissuemimetics" through the generation of polyvesicular arrays of defined architecture and to the control of their functional properties.
Decision support consists in helping a decision-maker to improve his/her decisions. However, clients requesting decision support are often themselves experts and are often taken by third parties ...and/or the general public to be responsible for the decisions they make. This predicament raises complex challenges for decision analysts, who have to avoid infringing upon the expertise and responsibility of the decision-maker. The case of diagnosis decision support in healthcare contexts is particularly illustrative. To support clinicians in their work and minimize the risk of medical error, various decision support systems have been developed, as part of information systems that are now ubiquitous in healthcare contexts. To develop, in collaboration with the hospitals of Lyon, a diagnostic decision support system for day-to-day customary consultations, we propose in this paper a critical analysis of current approaches to diagnostic decision support, which mainly consist in providing them with guidelines or even full-fledged diagnosis recommendations. We highlight that the use of such decision support systems by physicians raises responsibility issues, but also that it is at odds with the needs and constraints of customary consultations. We argue that the historical choice to favor guidelines or recommendations to physicians implies a very specific vision of what it means to support physicians, and we argue that the flaws of this vision partially explain why current diagnostic decision support systems are not accepted by physicians in their application to customary situations. Based on this analysis, we propose that decision support to physicians for customary cases should be deployed in an “adjustive” approach, which consists in providing physicians with the data on patients they need, when they need them, during consultations. The rationale articulated in this article has a more general bearing than clinical decision support and bears lessons for decision support activities in other contexts where decision-makers are competent and responsible experts.