Create Training Data Research Articles

Machine learning approaches for use in deblending

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.

Building-damage detection method based on machine learning utilizing aerial photographs of the Kumamoto earthquake

This article presents a method for detecting damaged buildings in the event of an earthquake using machine learning models and aerial photographs. We initially created training data for machine learning models using aerial photographs captured around the town of Mashiki immediately after the main shock of the 2016 Kumamoto earthquake. All buildings are classified into one of the four damage levels by visual interpretation. Subsequently, two damage discrimination models are developed: a bag-of-visual-words model and a model based on a convolutional neural network. Results are compared and validated in terms of accuracy, revealing that the latter model is preferable. Moreover, for the convolutional neural network model, the target areas are expanded and the recalls of damage classification at the four levels range approximately from 66% to 81%.

Event Based Sentiment Analysis on Futures Trading

Market expectations as well as perception of the investment risks and returns are dependent on information arrivals. News arrival forms the basis for market sentiment, which in turn forms the basis for trading positions. Research in sentiment analysis focuses on quantifying the impact that news has on prevailing market sentiment. However, it is not news but events that impact the market sentiment; and the news is one of the modes to disseminate information about the events. Sentiment analysis must distinguish the events from news and events should be used as the predicting construct for market sentiments. This paper proposes an event-based sentiment analysis model that entails event identification, event-based training data creation, and event representation algorithms. A comparative analysis of news-based and event-based sentiment analysis is done on high-frequency futures trades, using the real-time news as the source of market information. The proposed event-based sentiment analysis performed better than the traditional news-based sentiment analysis when evaluated using both the statistical metrics and simulated trading. This paper presents pivotal research in the direction of event-based sentiment analysis models and its implication on algorithmic trading.

Overall Methodology Design for the United States National Land Cover Database 2016 Products

The National Land Cover Database (NLCD) 2016 provides a suite of data products, including land cover and land cover change of the conterminous United States from 2001 to 2016, at two- to three-year intervals. The development of this product is part of an effort to meet the growing demand for longer temporal duration and more frequent, accurate, and consistent land cover and change information. To accomplish this, we designed a new land cover strategy and developed comprehensive methods, models, and procedures for NLCD 2016 implementation. Major steps in the new procedures consist of data preparation, land cover change detection and classification, theme-based postprocessing, and final integration. Data preparation includes Landsat imagery selection, cloud detection, and cloud filling, as well as compilation and creation of more than 30 national-scale ancillary datasets. Land cover change detection includes single-date water and snow/ice detection algorithms and models, two-date multi-index integrated change detection models, and long-term multi-date change algorithms and models. The land cover classification includes seven-date training data creation and 14-run classifications. Pools of training data for change and no-change areas were created before classification based on integrated information from ancillary data, change-detection results, Landsat spectral and temporal information, and knowledge-based trajectory analysis. In postprocessing, comprehensive models for each land cover theme were developed in a hierarchical order to ensure the spatial and temporal coherence of land cover and land cover changes over 15 years. An initial accuracy assessment on four selected Landsat path/rows classified with this method indicates an overall accuracy of 82.0% at an Anderson Level II classification and 86.6% at the Anderson Level I classification after combining the primary and alternate reference labels. This methodology was used for the operational production of NLCD 2016 for the Conterminous United States, with final produced products available for free download.

Twitter Sentiment Recognition using Support Vector Machine

In this we explore the effectiveness of language features to identify Twitter messages ' feelings. We assess the utility of existing lexical tools as well as capturing features of informal and innovative language knowledge used in micro blogging. We take a supervised approach to the problem, but to create training data, we use existing hash tags in the Twitter data. We Using three separate Twitter messaging companies in our experiments. We use the hash tagged data set (HASH) for development and training, which we compile from the Edinburgh Twitter corpus, and the emoticon data set (EMOT) from the I Sieve Corporation (ISIEVE) for evaluation. Twitter contains huge amount of data . This data may be of different types such as structured data or unstructured data. So by using this data and Appling pre processing techniques we can be able to read the comments from the users. And also the comments will be classified into three categories. They are positive negative and also the neutral comments.Today they use the processing of natural language, information, and text interpretation to derive and classify text feeling into pos itive, negative, and neutral categories. We can also examine the utility of language features to identify Twitter mess ages ' feelings. In addition, state-of - the-art approaches take into consideration only the tweet to be classified when classifying the feeling; they ignore its context (i.e. related tweets).Since tweets are usually short and more ambiguous, however, it is sometimes not enough to consider only the current tweet for classification of sentiments.Informal and innovative microblogging language. We take a sup ervised approach to the problem, but to create training data, we use existing hashtags in the Twitter data.This paper also contrasts sentiment analysis approaches in evaluating political views using Naïve Bayes supervised machine learning algorithm which performs in better analysis compared to other techniques Paper

Learning to Drive: End-to-End Off-Road Path Prediction

Autonomous driving is a field currently gaining a lot of attention, and recently ?end to end? approaches, whereby a machine learning algorithm learns to drive by emulating a human driver, have demonstrated significant potential. However, recent work has focused on the on-road environment, rather than the more challenging off-road environment. In this work we propose a new approach to this problem, whereby instead of learning to predict immediate driver control inputs, we train a deep convolutional neural network (CNN) to predict the future path that a vehicle will take through an off-road environment visually, addressing several limitations inherent in existing methods. We combine a novel approach to automatic training data creation, making use of stereoscopic visual odometry, with a state-of-the-art CNN architecture to map a predicted route directly onto image pixels, and demonstrate the effectiveness of our approach using our own off-road data set.

Application of Deep Neural Networks in the Problem of Obtaining Depth Maps from Two-Dimensional Images

Purpose of research is to study approaches to the depth map generation for deep neural networks testing and learning. The problem of obtaining information about the distance from the camera to the scenery object using a 2D image by means of deep neural networks without applying a stereocamera is considered. Methods. Generation of 3D scenery for training and assessment of the neural network was carried out using the 3D-computer graphics application Blender. The standard deviation (RMS) was used to estimate the accuracy of learning. Machine learning was implemented using the Keras library and optimization was implemented using the AdaGrad approach. Results . The architecture of a deep neural network which receives three sequences of 2D images from the 3D scenery video stream in the input and outputs the predicted depth map for the considered 3D scenery, is provided. The method for creating training data sets containing information about the depth of the map using Blender software is described. The problem of overtraining involving the fact that the created models work using specially generated data sets but still can not predict the correct depth map for random images is studied. The results of the testing actual methods for depth maps creation using deep neural networks are presented. Conclusion . The main problem of the proposed method is overtraining which can be expressed in predicting a certain average value for different images or predicting the same output for different inputs. To solve this problem, we can use already trained networks or training and variation samples containing 2D images of different sceneries.

Application of machine learning techniques in mineral phase segmentation for X-ray microcomputed tomography (µCT) data

X-ray microcomputed tomography (µCT) offers a non-destructive three-dimensional analysis of ores but its application in mineralogical analysis and mineral segmentation is relatively limited. In this study, the application of machine learning techniques for segmenting mineral phases in a µCT dataset is presented. Various techniques were implemented, including unsupervised classification as well as grayscale-based and feature-based supervised classification. A feature matching method was used to register the back-scattered electron (BSE) mineral map to its corresponding µCT slice, allowing automatic annotation of minerals in the µCT slice to create training data for the classifiers. Unsupervised classification produced satisfactory results in terms of segmenting between amphibole, plagioclase, and sulfide phases. However, the technique was not able to differentiate between sulfide phases in the case of chalcopyrite and pyrite. Using supervised classification, around 50–60% of the chalcopyrite and 97–99% of pyrite were correctly identified. Feature based classification was found to have a poorer sensitivity to chalcopyrite, but produced a better result in segmenting between the mineral grains, as it operates based on voxel regions instead of individual voxels. The mineralogical results from the 3D µCT data showed considerable difference compared to the BSE mineral map, indicating stereological error exhibited in the latter analysis. The main limitation of this approach lies in the dataset itself, in which there was a significant overlap in grayscale values between chalcopyrite and pyrite, therefore highly limiting the classifier accuracy.

3-D Point Cloud Registration Using Convolutional Neural Networks

This paper develops a registration architecture for the purpose of estimating relative pose including the rotation and the translation of an object in terms of a model in 3-D space based on 3-D point clouds captured by a 3-D camera. Particularly, this paper addresses the time-consuming problem of 3-D point cloud registration which is essential for the closed-loop industrial automated assembly systems that demand fixed time for accurate pose estimation. Firstly, two different descriptors are developed in order to extract coarse and detailed features of these point cloud data sets for the purpose of creating training data sets according to diversified orientations. Secondly, in order to guarantee fast pose estimation in fixed time, a seemingly novel registration architecture by employing two consecutive convolutional neural network (CNN) models is proposed. After training, the proposed CNN architecture can estimate the rotation between the model point cloud and a data point cloud, followed by the translation estimation based on computing average values. By covering a smaller range of uncertainty of the orientation compared with a full range of uncertainty covered by the first CNN model, the second CNN model can precisely estimate the orientation of the 3-D point cloud. Finally, the performance of the algorithm proposed in this paper has been validated by experiments in comparison with baseline methods. Based on these results, the proposed algorithm significantly reduces the estimation time while maintaining high precision.

Snorkel: rapid training data creation with weak supervision

Labeling training data is increasingly the largest bottleneck in deploying machine learning systems. We present Snorkel, a first-of-its-kind system that enables users to train state-of-the-art models without hand labeling any training data. Instead, users write labeling functions that express arbitrary heuristics, which can have unknown accuracies and correlations. Snorkel denoises their outputs without access to ground truth by incorporating the first end-to-end implementation of our recently proposed machine learning paradigm, data programming. We present a flexible interface layer for writing labeling functions based on our experience over the past year collaborating with companies, agencies, and research laboratories. In a user study, subject matter experts build models 2.8times faster and increase predictive performance an average 45.5% versus seven hours of hand labeling. We study the modeling trade-offs in this new setting and propose an optimizer for automating trade-off decisions that gives up to 1.8times speedup per pipeline execution. In two collaborations, with the US Department of Veterans Affairs and the US Food and Drug Administration, and on four open-source text and image data sets representative of other deployments, Snorkel provides 132% average improvements to predictive performance over prior heuristic approaches and comes within an average 3.60% of the predictive performance of large hand-curated training sets.

Landscape characteristics of non‐native pine plantations and invasions in Southern Chile

AbstractThe spread of non‐native conifers into areas naturally dominated by other vegetation types is a growing problem in South America. This process results in a landscape transformation as the conifers suppress native vegetation leading to reduced biodiversity, lower water availability and altered nutrient dynamics. Previous research highlights the broad spatial extents of land cover change in parts of Chile. However, in Southern Chile, the extent of plantations and the landscape characteristics associated with plantations and ongoing pine invasions are poorly understood. Here, we characterised non‐native pine land cover within one Landsat scene (World Reference System 2 Path 232/Row 92; ~34 000 km2) in Southern Chile. We created training data based on historical high‐resolution imagery, derived land cover predictors from time series of Landsat observations and used a Random Forest classifier to map the distribution of non‐native pines. The overall classification accuracy was 88%, and the accuracy of the non‐native pine class exceeded 90%. Although 71% of non‐native pine patches were within 500 m of other non‐native pine patches, isolated non‐native pine patches were found to occur up to 55 km from the nearest neighbour. These distant plantations could exacerbate invasion risk by creating propagule sources for novel invasion fronts. In relation to landscape characteristics, non‐native pines were found to be more likely to occur in low slope and mid‐elevation areas. Because most of the study area is native forest, most non‐native pine patches border native forest. However, non‐native pine patches were almost three times more likely than random patches to border grass/agriculture. This suggests that grasslands and disturbed sites, which have low resistance to non‐native pine invasion, are disproportionately exposed to pine propagules. Our results indicate that non‐native pine plantations are extensive across Southern Chile, and well poised to cause future invasion.

MOBILE LASER SCAN DATA FOR ROAD SURFACE DAMAGE DETECTION

Abstract. Road surface anomalies affect driving conditions, such as driving comfort and safety. Examples for such anomalies are potholes, cracks and ravelling. Automatic detection and localisation of these anomalies can be used for targeted road maintenance. Currently road damage is detected by road inspectors who drive slowly on the road to look out for surface anomalies, which can be dangerous. For improving the safety road inspectors can evaluate road images. However, results may be different as this evaluation is subjective. In this research a method is created for detecting road damage by using mobile profile laser scan data. First features are created, based on a sliding window. Then K-means clustering is used to create training data for a Random Forest algorithm. Finally, mathematical morphological operations are used to clean the data and connect the damage points. The result is an objective and detailed damage classification. The method is tested on a 120 meters long road data set that includes different types of damage. Validation is done by comparing the results to a classification of a human road inspector. However, the damage classification of the proposed method contains more details which makes validation difficult. Nevertheless does this method result in 79% overlap with the validation data. Although the results are already promising, developments such as pre-processing the data could lead to improvements.

A comparison of deep learning and citizen science techniques for counting wildlife in aerial survey images

Abstract Fast and accurate estimates of wildlife abundance are an essential component of efforts to conserve ecosystems in the face of rapid environmental change. A widely used method for estimating species abundance involves flying aerial transects, taking photographs, counting animals within the images and then inferring total population size based on a statistical estimate of species density in the region. The intermediate task of manually counting the aerial images is highly labour intensive and is often the limiting step in making a population estimate. Here, we assess the use of two novel approaches to perform this task by deploying both citizen scientists and deep learning to count aerial images of the 2015 survey of wildebeest (Connochaetes taurinus) in Serengeti National Park, Tanzania. Through the use of the online platform Zooniverse, we collected multiple non‐expert counts by citizen scientists and used three different aggregation methods to obtain a single count for the survey images. We also counted the images by developing a bespoke deep learning method via the use of a convolutional neural network. The results of both approaches were then compared. After filtering of the citizen science counts, both approaches provided highly accurate total estimates. The deep learning method was far faster and appears to be a more reliable and predictable approach; however, we note that citizen science volunteers played an important role when creating training data for the algorithm. Notably, our results show that accurate, species‐specific, automated counting of aerial wildlife images is now possible.

Selecting optimal bands for sub‐pixel target detection in hyperspectral images based on implanting synthetic targets

Target detection at sub-pixel abundances is, in fact, one of the challenging issues of hyperspectral image processing. Selection of optimal bands to improve sub-pixel target detection (STD) performance is one of the common solutions, applied by many researchers. Nevertheless, the absence of sufficient training data is the main weakness of selecting optimal bands with regard to this approach. The present research introduces a new band selection method for STD in hyperspectral images, based on creating training data, in which the desired target spectrum is implanted randomly in a series of host pixels from the entire hyperspectral image. Afterwards, via running an optimisation algorithm twice, with the aim of minimising the false alarm rate (FAR) in local adaptive coherence estimator target detection algorithm, the number of optimal bands and optimal spectral bands are selected. In this study, the performance of three optimisation methods including the genetic algorithm (GA), Grey Wolf optimisation (GWO), and particle swarm optimisation (PSO) are compared. Experimental results on HyMap and Hyperion datasets show that the proposed method obtains the minimum FAR compared with the rest of the evaluated methods. Also, based on the results obtained, GWO outperforms GA and PSO optimisation methods in the STD domain.

Machines as Craftsmen: Localized Parameter Setting Optimization for Fused Filament Fabrication 3D Printing

AbstractQuality control and repeatability of 3D printing must be enhanced to fully unlock its utility beyond prototyping and noncritical applications. Machine learning is a potential solution to improving 3D printing performance and is explored for areas including flaw identification and property prediction. However, critical problems must be resolved before machine learning can truly enable 3D printing to reach its potential, including the very large data sets required for training and the inherently local nature of 3D printing where the optimum parameter settings vary throughout the part. This work outlines an end‐to‐end tool for integrating machine learning into the 3D printing process. The tool selects the ideal parameter settings at each location, taking into consideration factors such as geometry, hardware and material response times, and operator priorities. The tool demonstrates its usefulness by correcting for visual flaws common in fused filament fabrication parts. An image recognition neural network classifies local flaws in parts to create training data. A gradient boosting classifier then predicts the local flaws in future parts, based on location, geometry, and parameter settings. The tool selects optimum parameter settings based on the aforementioned factors. The resulting prints show increased quality over prints that use global parameters only.

Domain-specific data augmentation for segmenting MR images of fatty infiltrated human thighs with neural networks.

Fat-fraction has been established as a relevant marker for the assessment and diagnosis of neuromuscular diseases. For computing this metric, segmentation of muscle tissue in MR images is a first crucial step. To tackle the high degree of variability in combination with the high annotation effort for training supervised segmentation models (such as fully convolutional neural networks). Prospective. In all, 41 patients consisting of 20 patients showing fatty infiltration and 21 healthy subjects. Field Strength/Sequence: The T1 -weighted MR-pulse sequences were acquired on a 1.5T scanner. To increase performance with limited training data, we propose a domain-specific technique for simulating fatty infiltrations (i.e., texture augmentation) in nonaffected subjects' MR images in combination with shape augmentation. For simulating the fatty infiltrations, we make use of an architecture comprising several competing networks (generative adversarial networks) that facilitate a realistic artificial conversion between healthy and infiltrated MR images. Finally, we assess the segmentation accuracy (Dice similarity coefficient). A Wilcoxon signed rank test was performed to assess whether differences in segmentation accuracy are significant. The mean Dice similarity coefficients significantly increase from 0.84-0.88 (P < 0.01) using data augmentation if training is performed with mixed data and from 0.59-0.87 (P < 0.001) if training is conducted with healthy subjects only. Domain-specific data adaptation is highly suitable for facilitating neural network-based segmentation of thighs with feasible manual effort for creating training data. The results even suggest an approach completely bypassing manual annotations. 4 Technical Efficacy: Stage 3.

A clinical text classification paradigm using weak supervision and deep representation

BackgroundAutomatic clinical text classification is a natural language processing (NLP) technology that unlocks information embedded in clinical narratives. Machine learning approaches have been shown to be effective for clinical text classification tasks. However, a successful machine learning model usually requires extensive human efforts to create labeled training data and conduct feature engineering. In this study, we propose a clinical text classification paradigm using weak supervision and deep representation to reduce these human efforts.MethodsWe develop a rule-based NLP algorithm to automatically generate labels for the training data, and then use the pre-trained word embeddings as deep representation features for training machine learning models. Since machine learning is trained on labels generated by the automatic NLP algorithm, this training process is called weak supervision. We evaluat the paradigm effectiveness on two institutional case studies at Mayo Clinic: smoking status classification and proximal femur (hip) fracture classification, and one case study using a public dataset: the i2b2 2006 smoking status classification shared task. We test four widely used machine learning models, namely, Support Vector Machine (SVM), Random Forest (RF), Multilayer Perceptron Neural Networks (MLPNN), and Convolutional Neural Networks (CNN), using this paradigm. Precision, recall, and F1 score are used as metrics to evaluate performance.ResultsCNN achieves the best performance in both institutional tasks (F1 score: 0.92 for Mayo Clinic smoking status classification and 0.97 for fracture classification). We show that word embeddings significantly outperform tf-idf and topic modeling features in the paradigm, and that CNN captures additional patterns from the weak supervision compared to the rule-based NLP algorithms. We also observe two drawbacks of the proposed paradigm that CNN is more sensitive to the size of training data, and that the proposed paradigm might not be effective for complex multiclass classification tasks.ConclusionThe proposed clinical text classification paradigm could reduce human efforts of labeled training data creation and feature engineering for applying machine learning to clinical text classification by leveraging weak supervision and deep representation. The experimental experiments have validated the effectiveness of paradigm by two institutional and one shared clinical text classification tasks.

Bootstrapping a Neural Morphological Analyzer for St. Lawrence Island Yupik from a Finite-State Transducer

Morphological analysis is a critical enabling technology for polysynthetic languages. We present a neural morphological analyzer for case-inflected nouns in St. Lawrence Island Yupik, an endangered polysythetic language in the Inuit-Yupik language family, treating morphological analysis as a recurrent neural sequence-to-sequence task. By utilizing an existing finite-state morphological analyzer to create training data, we improve analysis coverage on attested Yupik word types from approximately 75% for the existing finite-state analyzer to 100% for the neural analyzer. At the same time, we achieve a substantially higher level of accuracy on a held-out testing set, from 78.9% accuracy for the finite-state analyzer to 92.2% accuracy for our neural analyzer.

Real-time data-driven interactive rough sketch inking

We present an interactive approach for inking , which is the process of turning a pencil rough sketch into a clean line drawing. The approach, which we call the Smart Inker , consists of several "smart" tools that intuitively react to user input, while guided by the input rough sketch, to efficiently and naturally connect lines, erase shading, and fine-tune the line drawing output. Our approach is data-driven: the tools are based on fully convolutional networks, which we train to exploit both the user edits and inaccurate rough sketch to produce accurate line drawings, allowing high-performance interactive editing in real-time on a variety of challenging rough sketch images. For the training of the tools, we developed two key techniques: one is the creation of training data by simulation of vague and quick user edits; the other is a line normalization based on learning from vector data. These techniques, in combination with our sketch-specific data augmentation, allow us to train the tools on heterogeneous data without actual user interaction. We validate our approach with an in-depth user study, comparing it with professional illustration software, and show that our approach is able to reduce inking time by a factor of 1.8X, while improving the results of amateur users.

ROLE OF VARIOUS DATA MINING TECHNIQUES IN SENTIMENTAL ANALYSIS

Data is present in large amount and it became quite hectic to deal with with this amount of data but data mining has made it easy .Data mining has given various concepts which has made the process of sentimental analysis easy . Sentiment analysis involves the usage of text scrutiny and processing in order to recognize the patterns plus gaining the useful information. Various techniques namely Natural Language Processing, Machine learning, Text mining are involved in this process of sentiment analysis. By collecting different opinions, polarity is find out and is categorized as positive, negative or neutral. :1.0pt 0in 1.0pt 0in'> Rapid expansion of cloud technologies were mainly due to the increased requirements of cloud users. However, increased requests also laden with increased resource requirements especially due to the elastic nature of the cloud. This mandates the need for effective resource provisioning model. This paper presents a Time Window based Auto-Regressive Hybrid PSO (TWARP) model that provides faster and more appropriate resource allocations. The TWARP model is composed of a temporal data grouping model to create training data, an auto-regression model to predict future requirements, a PSO-SA based optimal package selection mechanism and a final request handling mechanism that allocates the actual resource to a user. Experiments indicate low time requirements and effective allocation levels. Comparison with recent literature works also indicates highly effective performances of the proposed model.