SEMCITY TOULOUSE: A BENCHMARK FOR BUILDING INSTANCE SEGMENTATION IN SATELLITE IMAGES

Abstract. In order to reach the goal of reliably solving Earth monitoring tasks, automated and efficient machine learning methods are necessary for large-scale scene analysis and interpretation. A typical bottleneck of supervised learning approaches is the availability of accurate (manually) labeled training data, which is particularly important to train state-of-the-art (deep) learning methods. We present SemCity Toulouse, a publicly available, very high resolution, multi-spectral benchmark data set for training and evaluation of sophisticated machine learning models. The benchmark acts as test bed for single building instance segmentation which has been rarely considered before in densely built urban areas. Additional information is provided in the form of a multi-class semantic segmentation annotation covering the same area plus an adjacent area 3 times larger. The data set addresses interested researchers from various communities such as photogrammetry and remote sensing, but also computer vision and machine learning.

Download Full-text

Machine learning identifies an immunological pattern associated with multiple juvenile idiopathic arthritis subtypes

Annals of the Rheumatic Diseases ◽

10.1136/annrheumdis-2018-214354 ◽

2019 ◽

Vol 78 (5) ◽

pp. 617-628 ◽

Cited By ~ 5

Author(s):

Erika Van Nieuwenhove ◽

Vasiliki Lagou ◽

Lien Van Eyck ◽

James Dooley ◽

Ulrich Bodenhofer ◽

...

Keyword(s):

Machine Learning ◽

Juvenile Idiopathic Arthritis ◽

Large Scale ◽

Inflammatory Diseases ◽

Adaptive Immune System ◽

Healthy Children ◽

Learning Approaches ◽

Data Set ◽

Immune Signature ◽

Systemic Jia

ObjectivesJuvenile idiopathic arthritis (JIA) is the most common class of childhood rheumatic diseases, with distinct disease subsets that may have diverging pathophysiological origins. Both adaptive and innate immune processes have been proposed as primary drivers, which may account for the observed clinical heterogeneity, but few high-depth studies have been performed.MethodsHere we profiled the adaptive immune system of 85 patients with JIA and 43 age-matched controls with indepth flow cytometry and machine learning approaches.ResultsImmune profiling identified immunological changes in patients with JIA. This immune signature was shared across a broad spectrum of childhood inflammatory diseases. The immune signature was identified in clinically distinct subsets of JIA, but was accentuated in patients with systemic JIA and those patients with active disease. Despite the extensive overlap in the immunological spectrum exhibited by healthy children and patients with JIA, machine learning analysis of the data set proved capable of discriminating patients with JIA from healthy controls with ~90% accuracy.ConclusionsThese results pave the way for large-scale immune phenotyping longitudinal studies of JIA. The ability to discriminate between patients with JIA and healthy individuals provides proof of principle for the use of machine learning to identify immune signatures that are predictive to treatment response group.

Download Full-text

Prediction of Collapsibility of Loess of Construction Sites in Xining Based on Machine Learning Methods

10.21203/rs.3.rs-307514/v1 ◽

2021 ◽

Author(s):

Qifei Zhao ◽

Xiaojun Li ◽

Yunning Cao ◽

Zhikun Li ◽

Jixin Fan

Keyword(s):

Machine Learning ◽

Support Vector Machine ◽

Training Data ◽

Support Vector ◽

Engineering Practice ◽

Burial Depth ◽

Learning Methods ◽

Data Set ◽

Machine Learning Methods ◽

North East

Abstract Collapsibility of loess is a significant factor affecting engineering construction in loess area, and testing the collapsibility of loess is costly. In this study, A total of 4,256 loess samples are collected from the north, east, west and middle regions of Xining. 70% of the samples are used to generate training data set, and the rest are used to generate verification data set, so as to construct and validate the machine learning models. The most important six factors are selected from thirteen factors by using Grey Relational analysis and multicollinearity analysis: burial depth、water content、specific gravity of soil particles、void rate、geostatic stress and plasticity limit. In order to predict the collapsibility of loess, four machine learning methods: Support Vector Machine (SVM), Random Subspace Based Support Vector Machine (RSSVM), Random Forest (RF) and Naïve Bayes Tree (NBTree), are studied and compared. The receiver operating characteristic (ROC) curve indicators, standard error (SD) and 95% confidence interval (CI) are used to verify and compare the models in different research areas. The results show that: RF model is the most efficient in predicting the collapsibility of loess in Xining, and its AUC average is above 80%, which can be used in engineering practice.

Download Full-text

Influence of synchronization within a sensor network on machine learning results

Journal of Sensors and Sensor Systems ◽

10.5194/jsss-10-233-2021 ◽

2021 ◽

Vol 10 (2) ◽

pp. 233-245

Author(s):

Tanja Dorst ◽

Yannick Robin ◽

Sascha Eichstädt ◽

Andreas Schütze ◽

Tizian Schneider

Keyword(s):

Machine Learning ◽

Feature Extraction ◽

Data Augmentation ◽

Time Synchronization ◽

Fundamental Problem ◽

Training Data ◽

Sensor Data ◽

Test Bed ◽

Data Set ◽

Linear Discriminant

Abstract. Process sensor data allow for not only the control of industrial processes but also an assessment of plant conditions to detect fault conditions and wear by using sensor fusion and machine learning (ML). A fundamental problem is the data quality, which is limited, inter alia, by time synchronization problems. To examine the influence of time synchronization within a distributed sensor system on the prediction performance, a test bed for end-of-line tests, lifetime prediction, and condition monitoring of electromechanical cylinders is considered. The test bed drives the cylinder in a periodic cycle at maximum load, a 1 s period at constant drive speed is used to predict the remaining useful lifetime (RUL). The various sensors for vibration, force, etc. integrated into the test bed are sampled at rates between 10 kHz and 1 MHz. The sensor data are used to train a classification ML model to predict the RUL with a resolution of 1 % based on feature extraction, feature selection, and linear discriminant analysis (LDA) projection. In this contribution, artificial time shifts of up to 50 ms between individual sensors' cycles are introduced, and their influence on the performance of the RUL prediction is investigated. While the ML model achieves good results if no time shifts are introduced, we observed that applying the model trained with unmodified data only to data sets with time shifts results in very poor performance of the RUL prediction even for small time shifts of 0.1 ms. To achieve an acceptable performance also for time-shifted data and thus achieve a more robust model for application, different approaches were investigated. One approach is based on a modified feature extraction approach excluding the phase values after Fourier transformation; a second is based on extending the training data set by including artificially time-shifted data. This latter approach is thus similar to data augmentation used to improve training of neural networks.

Download Full-text

Machine-learning methods in the classification of water bodies

Environmental & Socio-economic Studies ◽

10.1515/environ-2016-0010 ◽

2016 ◽

Vol 4 (2) ◽

pp. 34-42 ◽

Cited By ~ 1

Author(s):

Marek Sołtysiak ◽

Marcin Blachnik ◽

Dominika Dąbrowska

Keyword(s):

Machine Learning ◽

Water Body ◽

Urban Areas ◽

Water Bodies ◽

Learning Methods ◽

Amphibian Species ◽

Data Set ◽

Machine Learning Methods ◽

Nearest Neighbours

AbstractAmphibian species have been considered as useful ecological indicators. They are used as indicators of environmental contamination, ecosystem health and habitat quality., Amphibian species are sensitive to changes in the aquatic environment and therefore, may form the basis for the classification of water bodies. Water bodies in which there are a large number of amphibian species are especially valuable even if they are located in urban areas. The automation of the classification process allows for a faster evaluation of the presence of amphibian species in the water bodies. Three machine-learning methods (artificial neural networks, decision trees and the k-nearest neighbours algorithm) have been used to classify water bodies in Chorzów – one of 19 cities in the Upper Silesia Agglomeration. In this case, classification is a supervised data mining method consisting of several stages such as building the model, the testing phase and the prediction. Seven natural and anthropogenic features of water bodies (e.g. the type of water body, aquatic plants, the purpose of the water body (destination), position of the water body in relation to any possible buildings, condition of the water body, the degree of littering, the shore type and fishing activities) have been taken into account in the classification. The data set used in this study involved information about 71 different water bodies and 9 amphibian species living in them. The results showed that the best average classification accuracy was obtained with the multilayer perceptron neural network.

Download Full-text

A Comparison of Machine Learning Approaches to Improve Free Topography Data for Flood Modelling

Remote Sensing ◽

10.3390/rs13020275 ◽

2021 ◽

Vol 13 (2) ◽

pp. 275

Author(s):

Michael Meadows ◽

Matthew Wilson

Keyword(s):

Neural Network ◽

Machine Learning ◽

Spatial Patterns ◽

Large Scale ◽

Multiple Scales ◽

Flood Hazard ◽

Training Data ◽

Learning Approaches ◽

Testing Dataset ◽

Topography Data

Given the high financial and institutional cost of collecting and processing accurate topography data, many large-scale flood hazard assessments continue to rely instead on freely-available global Digital Elevation Models, despite the significant vertical biases known to affect them. To predict (and thereby reduce) these biases, we apply a fully-convolutional neural network (FCN), a form of artificial neural network originally developed for image segmentation which is capable of learning from multi-variate spatial patterns at different scales. We assess its potential by training such a model on a wide variety of remote-sensed input data (primarily multi-spectral imagery), using high-resolution, LiDAR-derived Digital Terrain Models published by the New Zealand government as the reference topography data. In parallel, two more widely used machine learning models are also trained, in order to provide benchmarks against which the novel FCN may be assessed. We find that the FCN outperforms the other models (reducing root mean square error in the testing dataset by 71%), likely due to its ability to learn from spatial patterns at multiple scales, rather than only a pixel-by-pixel basis. Significantly for flood hazard modelling applications, corrections were found to be especially effective along rivers and their floodplains. However, our results also suggest that models are likely to be biased towards the land cover and relief conditions most prevalent in their training data, with further work required to assess the importance of limiting training data inputs to those most representative of the intended application area(s).

Download Full-text

Targeting Prospective Customers: Robustness of Machine-Learning Methods to Typical Data Challenges

Management Science ◽

10.1287/mnsc.2019.3308 ◽

2020 ◽

Vol 66 (6) ◽

pp. 2495-2522 ◽

Cited By ~ 2

Author(s):

Duncan Simester ◽

Artem Timoshenko ◽

Spyros I. Zoumpoulis

Keyword(s):

Machine Learning ◽

Field Experiment ◽

Large Scale ◽

Field Experiments ◽

Training Data ◽

Classification Methods ◽

Learning Methods ◽

Model Driven ◽

Machine Learning Methods

We investigate how firms can use the results of field experiments to optimize the targeting of promotions when prospecting for new customers. We evaluate seven widely used machine-learning methods using a series of two large-scale field experiments. The first field experiment generates a common pool of training data for each of the seven methods. We then validate the seven optimized policies provided by each method together with uniform benchmark policies in a second field experiment. The findings not only compare the performance of the targeting methods, but also demonstrate how well the methods address common data challenges. Our results reveal that when the training data are ideal, model-driven methods perform better than distance-driven methods and classification methods. However, the performance advantage vanishes in the presence of challenges that affect the quality of the training data, including the extent to which the training data captures details of the implementation setting. The challenges we study are covariate shift, concept shift, information loss through aggregation, and imbalanced data. Intuitively, the model-driven methods make better use of the information available in the training data, but the performance of these methods is more sensitive to deterioration in the quality of this information. The classification methods we tested performed relatively poorly. We explain the poor performance of the classification methods in our setting and describe how the performance of these methods could be improved. This paper was accepted by Matthew Shum, marketing.

Download Full-text

Fault Diagnosis via Neural Ordinary Differential Equations

Applied Sciences ◽

10.3390/app11093776 ◽

2021 ◽

Vol 11 (9) ◽

pp. 3776

Author(s):

Luis Enciso-Salas ◽

Gustavo Pérez-Zuñiga ◽

Javier Sotomayor-Moriano

Keyword(s):

Machine Learning ◽

Fault Diagnosis ◽

Differential Equations ◽

Ordinary Differential Equations ◽

Data Augmentation ◽

Hybrid Approach ◽

Training Data ◽

Data Series ◽

Learning Approaches ◽

Data Set

Implementation of model-based fault diagnosis systems can be a difficult task due to the complex dynamics of most systems, an appealing alternative to avoiding modeling is to use machine learning-based techniques for which the implementation is more affordable nowadays. However, the latter approach often requires extensive data processing. In this paper, a hybrid approach using recent developments in neural ordinary differential equations is proposed. This approach enables us to combine a natural deep learning technique with an estimated model of the system, making the training simpler and more efficient. For evaluation of this methodology, a nonlinear benchmark system is used by simulation of faults in actuators, sensors, and process. Simulation results show that the proposed methodology requires less processing for the training in comparison with conventional machine learning approaches since the data-set is directly taken from the measurements and inputs. Furthermore, since the model used in the essay is only a structural approximation of the plant; no advanced modeling is required. This approach can also alleviate some pitfalls of training data-series, such as complicated data augmentation methodologies and the necessity for big amounts of data.

Download Full-text

Improving stock trading decisions based on pattern recognition using machine learning technology

PLoS ONE ◽

10.1371/journal.pone.0255558 ◽

2021 ◽

Vol 16 (8) ◽

pp. e0255558

Author(s):

Yaohu Lin ◽

Shancun Liu ◽

Haijun Yang ◽

Harris Wu ◽

Bingbing Jiang

Keyword(s):

Machine Learning ◽

Pattern Recognition ◽

Investment Strategy ◽

Training Data ◽

Stock Trading ◽

Learning Methods ◽

Data Set ◽

Empirical Results ◽

Machine Learning Methods ◽

Trading Decisions

PRML, a novel candlestick pattern recognition model using machine learning methods, is proposed to improve stock trading decisions. Four popular machine learning methods and 11 different features types are applied to all possible combinations of daily patterns to start the pattern recognition schedule. Different time windows from one to ten days are used to detect the prediction effect at different periods. An investment strategy is constructed according to the identified candlestick patterns and suitable time window. We deploy PRML for the forecast of all Chinese market stocks from Jan 1, 2000 until Oct 30, 2020. Among them, the data from Jan 1, 2000 to Dec 31, 2014 is used as the training data set, and the data set from Jan 1, 2015 to Oct 30, 2020 is used to verify the forecasting effect. Empirical results show that the two-day candlestick patterns after filtering have the best prediction effect when forecasting one day ahead; these patterns obtain an average annual return, an annual Sharpe ratio, and an information ratio as high as 36.73%, 0.81, and 2.37, respectively. After screening, three-day candlestick patterns also present a beneficial effect when forecasting one day ahead in that these patterns show stable characteristics. Two other popular machine learning methods, multilayer perceptron network and long short-term memory neural networks, are applied to the pattern recognition framework to evaluate the dependency of the prediction model. A transaction cost of 0.2% is considered on the two-day patterns predicting one day ahead, thus confirming the profitability. Empirical results show that applying different machine learning methods to two-day and three-day patterns for one-day-ahead forecasts can be profitable.

Download Full-text

Machine learning approaches for use in deblending

The Leading Edge ◽

10.1190/tle39030188.1 ◽

2020 ◽

Vol 39 (3) ◽

pp. 188-194

Author(s):

Rolf H. Baardman ◽

Rob F. Hegge

Keyword(s):

Machine Learning ◽

A Priori ◽

Learning Algorithms ◽

Machine Learning Algorithms ◽

Training Data ◽

Learning Approaches ◽

Learning Methods ◽

Unseen Data ◽

Noise Classification ◽

Blended Data

Machine learning has grown into a topic of much interest in the seismic industry. Recently, machine learning was introduced in the field of seismic processing for applications such as demultiple, regularization, and tomography. Here, two novel machine learning algorithms are introduced that can perform deblending and automated blending noise classification. Conventional deblending algorithms require a priori information and user expertise to properly select and parameterize a specific algorithm. The potential benefits of machine learning methods include their hands-off implementation and their ability to learn an efficient deblending algorithm directly from data. The introduced methods are supervised learning methods. Their specific tasks (deblending/noise classification) are learned from training data consisting of data example pairs of input and labeled output. For instance, training a deblending algorithm requires pairs of blended data with their unblended counterparts. The availability of training data or the possibility of creating training data are key to the success of these supervised methods. Another aspect is how well the algorithms generalize. Can we expect good performance on (unseen) data that vary from the training data? We address these aspects and further illustrate with synthetic and field data examples. The classification and deblending examples show promising results, indicating that these machine learning algorithms can support and/or replace existing deblending approaches.

Download Full-text

Recent Progress in Machine Learning-based Prediction of Peptide Activity for Drug Discovery

Current Topics in Medicinal Chemistry ◽

10.2174/1568026619666190122151634 ◽

2019 ◽

Vol 19 (1) ◽

pp. 4-16 ◽

Cited By ~ 6

Author(s):

Qihui Wu ◽

Hanzhong Ke ◽

Dongli Li ◽

Qi Wang ◽

Jiansong Fang ◽

...

Keyword(s):

Machine Learning ◽

Drug Discovery ◽

Large Scale ◽

Recent Progress ◽

High Specificity ◽

Learning Approaches ◽

Anticancer Peptides ◽

The Past ◽

Traditional Approaches ◽

Large Scale Screening

Over the past decades, peptide as a therapeutic candidate has received increasing attention in drug discovery, especially for antimicrobial peptides (AMPs), anticancer peptides (ACPs) and antiinflammatory peptides (AIPs). It is considered that the peptides can regulate various complex diseases which are previously untouchable. In recent years, the critical problem of antimicrobial resistance drives the pharmaceutical industry to look for new therapeutic agents. Compared to organic small drugs, peptide- based therapy exhibits high specificity and minimal toxicity. Thus, peptides are widely recruited in the design and discovery of new potent drugs. Currently, large-scale screening of peptide activity with traditional approaches is costly, time-consuming and labor-intensive. Hence, in silico methods, mainly machine learning approaches, for their accuracy and effectiveness, have been introduced to predict the peptide activity. In this review, we document the recent progress in machine learning-based prediction of peptides which will be of great benefit to the discovery of potential active AMPs, ACPs and AIPs.

Download Full-text