Neural Supermodeling

2020 ◽  
Author(s):  
Wim Wiegerinck
Keyword(s):  
Deep Learning ◽  
Domain Knowledge ◽  
Model Performance ◽  
Ground Truth ◽  
Training Data ◽  
Combination Method ◽  
Neural Network Approach ◽  
Modeling Approach ◽  
Promising Tool ◽  
Spatially Extended

<p>Deep learning is a modeling approach that has shown impressive results in image processing and is arguably a promising tool for dealing with spatially extended complex systems such earth atmosphere with its visually interpretable patterns. A disadvantage of the neural network approach is that it typically requires an enormous amount of training data.</p><p> </p><p>Another recently proposed modeling approach is supermodeling. In supermodeling it is assumed that a dynamical system – the truth – is modelled by a set of good but imperfect models. The idea is to improve model performance by dynamically combining imperfect models during the simulation. The resulting combination of models is called the supermodel. The combination strength has to be learned from data. However, since supermodels do not start from scratch, but make use of existing domain knowledge, they may learn from less data.</p><p> </p><p>One of the ways to combine models is to define the tendencies of the supermodel as linear (weighted) combinations of the imperfect model tendencies. Several methods including linear regression have been proposed to optimize the weights.  However, the combination method might also be nonlinear. In this work we propose and explore a novel combination of deep learning and supermodeling, in which convolutional neural networks are used as tool to combine the predictions of the imperfect models.  The different supermodeling strategies are applied in simulations in a controlled environment with a three-level, quasi-geostrophic spectral model that serves as ground truth and perturbed models that serve as the imperfect models.</p>

scholarly journals USING SEMANTICALLY PAIRED IMAGES TO IMPROVE DOMAIN ADAPTATION FOR THE SEMANTIC SEGMENTATION OF AERIAL IMAGES

2020 ◽  
Vol V-2-2020 ◽  
pp. 483-492
Author(s):  
D. Gritzner ◽  
J. Ostermann
Keyword(s):  
Time Window ◽  
Domain Adaptation ◽  
Geographical Area ◽  
Model Performance ◽  
Ground Truth ◽  
Semantic Segmentation ◽  
Training Data ◽  
Aerial Images ◽  
Target Domain ◽  
Training Examples

Abstract. Modern machine learning, especially deep learning, which is used in a variety of applications, requires a lot of labelled data for model training. Having an insufficient amount of training examples leads to models which do not generalize well to new input instances. This is a particular significant problem for tasks involving aerial images: often training data is only available for a limited geographical area and a narrow time window, thus leading to models which perform poorly in different regions, at different times of day, or during different seasons. Domain adaptation can mitigate this issue by using labelled source domain training examples and unlabeled target domain images to train a model which performs well on both domains. Modern adversarial domain adaptation approaches use unpaired data. We propose using pairs of semantically similar images, i.e., whose segmentations are accurate predictions of each other, for improved model performance. In this paper we show that, as an upper limit based on ground truth, using semantically paired aerial images during training almost always increases model performance with an average improvement of 4.2% accuracy and .036 mean intersection-over-union (mIoU). Using a practical estimate of semantic similarity, we still achieve improvements in more than half of all cases, with average improvements of 2.5% accuracy and .017 mIoU in those cases.

scholarly journals On the objectivity, reliability, and validity of deep learning enabled bioimage analyses

eLife ◽  
2020 ◽  
Vol 9 ◽  
Author(s):  
Dennis Segebarth ◽  
Matthias Griebel ◽  
Nikolai Stein ◽  
Cora R von Collenberg ◽  
Corinna Martin ◽  
...  
Keyword(s):  
Deep Learning ◽  
Signal To Noise Ratio ◽  
Biological Effects ◽  
Reliability And Validity ◽  
Ground Truth ◽  
Training Data ◽  
Model Organisms ◽  
Data Annotation ◽  
Bioimage Analysis ◽  
Model Training

Bioimage analysis of fluorescent labels is widely used in the life sciences. Recent advances in deep learning (DL) allow automating time-consuming manual image analysis processes based on annotated training data. However, manual annotation of fluorescent features with a low signal-to-noise ratio is somewhat subjective. Training DL models on subjective annotations may be instable or yield biased models. In turn, these models may be unable to reliably detect biological effects. An analysis pipeline integrating data annotation, ground truth estimation, and model training can mitigate this risk. To evaluate this integrated process, we compared different DL-based analysis approaches. With data from two model organisms (mice, zebrafish) and five laboratories, we show that ground truth estimation from multiple human annotators helps to establish objectivity in fluorescent feature annotations. Furthermore, ensembles of multiple models trained on the estimated ground truth establish reliability and validity. Our research provides guidelines for reproducible DL-based bioimage analyses.

Deep learning-driven velocity model building workflow

2019 ◽  
Vol 38 (11) ◽  
pp. 872a1-872a9 ◽  
Author(s):  
Mauricio Araya-Polo ◽  
Stuart Farris ◽  
Manuel Florez
Keyword(s):  
Deep Learning ◽  
Seismic Data ◽  
Model Building ◽  
Ground Truth ◽  
Velocity Model ◽  
Training Data ◽  
Quality Data ◽  
Generative Adversarial Network ◽  
Data Set ◽  
Velocity Models

Exploration seismic data are heavily manipulated before human interpreters are able to extract meaningful information regarding subsurface structures. This manipulation adds modeling and human biases and is limited by methodological shortcomings. Alternatively, using seismic data directly is becoming possible thanks to deep learning (DL) techniques. A DL-based workflow is introduced that uses analog velocity models and realistic raw seismic waveforms as input and produces subsurface velocity models as output. When insufficient data are used for training, DL algorithms tend to overfit or fail. Gathering large amounts of labeled and standardized seismic data sets is not straightforward. This shortage of quality data is addressed by building a generative adversarial network (GAN) to augment the original training data set, which is then used by DL-driven seismic tomography as input. The DL tomographic operator predicts velocity models with high statistical and structural accuracy after being trained with GAN-generated velocity models. Beyond the field of exploration geophysics, the use of machine learning in earth science is challenged by the lack of labeled data or properly interpreted ground truth, since we seldom know what truly exists beneath the earth's surface. The unsupervised approach (using GANs to generate labeled data)illustrates a way to mitigate this problem and opens geology, geophysics, and planetary sciences to more DL applications.

scholarly journals Improving generalization of deep learning models for diagnostic pathology by increasing variability in training data: experiments on osteosarcoma subtypes

2020 ◽  
Author(s):  
Haiming Tang ◽  
Nanfei Sun ◽  
Steven Shen
Keyword(s):  
Deep Learning ◽  
Model Performance ◽  
High Variability ◽  
Training Data ◽  
Classification Model ◽  
Training Dataset ◽  
Learning Models ◽  
Diagnostic Pathology ◽  
Model Generalization ◽  
Histopathological Images

Artificial intelligence (AI) has an emerging progress in diagnostic pathology. A large number of studies of applying deep learning models to histopathological images have been published in recent years. While many studies claim high accuracies, they may fall into the pitfalls of overfitting and lack of generalization due to the high variability of the histopathological images. We use the example of Osteosarcoma to illustrate the pitfalls and how the addition of model input variability can help improve model performance. We use the publicly available osteosarcoma dataset to retrain a previously published classification model for osteosarcoma. We partition the same set of images into the training and testing datasets differently than the original study: the test dataset consists of images from one patient while the training dataset consists images of all other patients. The performance of the model on the test set using the new partition schema declines dramatically, indicating a lack of model generalization and overfitting.We also show the influence of training data variability on model performance by collecting a minimal dataset of 10 osteosarcoma subtypes as well as benign tissues and benign bone tumors of differentiation. We show the additions of more and more subtypes into the training data step by step under the same model schema yield a series of coherent models with increasing performances. In conclusion, we bring forward data preprocessing and collection tactics for histopathological images of high variability to avoid the pitfalls of overfitting and build deep learning models of higher generalization abilities.

scholarly journals EVICAN—a balanced dataset for algorithm development in cell and nucleus segmentation

2020 ◽  
Vol 36 (12) ◽  
pp. 3863-3870
Author(s):  
Mischa Schwendy ◽  
Ronald E Unger ◽  
Sapun H Parekh
Keyword(s):  
Deep Learning ◽  
Cell Biology ◽  
Ground Truth ◽  
Training Data ◽  
Training Dataset ◽  
Visual Cell ◽  
Application Development ◽  
Ground Truth Data ◽  
Quantitative Image ◽  
Nucleus Segmentation

Abstract Motivation Deep learning use for quantitative image analysis is exponentially increasing. However, training accurate, widely deployable deep learning algorithms requires a plethora of annotated (ground truth) data. Image collections must contain not only thousands of images to provide sufficient example objects (i.e. cells), but also contain an adequate degree of image heterogeneity. Results We present a new dataset, EVICAN—Expert visual cell annotation, comprising partially annotated grayscale images of 30 different cell lines from multiple microscopes, contrast mechanisms and magnifications that is readily usable as training data for computer vision applications. With 4600 images and ∼26 000 segmented cells, our collection offers an unparalleled heterogeneous training dataset for cell biology deep learning application development. Availability and implementation The dataset is freely available (https://edmond.mpdl.mpg.de/imeji/collection/l45s16atmi6Aa4sI?q=). Using a Mask R-CNN implementation, we demonstrate automated segmentation of cells and nuclei from brightfield images with a mean average precision of 61.6 % at a Jaccard Index above 0.5.

A Deep Neural Network Model for Packing Density Predictions and its Application in the Study of 1.5 Million Organic Molecules

2019 ◽  
Author(s):  
Mohammad Atif Faiz Afzal ◽  
Aditya Sonpal ◽  
Mojtaba Haghighatlari ◽  
Andrew J. Schultz ◽  
Johannes Hachmann
Keyword(s):  
Neural Network ◽  
High Throughput Screening ◽  
Organic Molecules ◽  
Model Performance ◽  
Training Data ◽  
Neural Network Approach ◽  
Massive Scale ◽  
Computationally Intensive ◽  
Computational Modeling And Simulation ◽  
Dynamics Simulations

<pre>The process of developing new compounds and materials is increasingly driven by computational modeling and simulation, which allow us to characterize candidates before pursuing them in the laboratory. One of the non-trivial properties of interest for organic materials is their packing in the bulk, which is highly dependent on their molecular structure. By controlling the latter, we can realize materials with a desired density (as well as other target properties). Molecular dynamics simulations are a popular and reasonably accurate way to compute the bulk density of molecules, however, since these calculations are computationally intensive, they are not a practically viable option for high-throughput screening studies that assess material candidates on a massive scale. In this work, we employ machine learning to develop a data-derived prediction model that is an alternative to physics-based simulations, and we utilize it for the hyperscreening of 1.5 million small organic molecules as well as to gain insights into the relationship between structural makeup and packing density.We also use this study to analyze the learning curve of the employed neural network approach and gain empirical data on the dependence of model performance and training data size, which will inform future investigations.</pre>

scholarly journals Deep learning approach for automatic landmark detection and alignment analysis in whole-spine lateral radiographs

2021 ◽  
Vol 11 (1) ◽  
Author(s):  
Yu-Cheng Yeh ◽  
Chi-Hung Weng ◽  
Yu-Jui Huang ◽  
Chen-Ju Fu ◽  
Tsung-Ting Tsai ◽  
...  
Keyword(s):  
Deep Learning ◽  
Model Performance ◽  
Ground Truth ◽  
Landmark Detection ◽  
Anatomic Landmarks ◽  
Radiographic Parameters ◽  
Learning Speed ◽  
Analysis System ◽  
Whole Spine ◽  
Alignment Analysis

AbstractHuman spinal balance assessment relies considerably on sagittal radiographic parameter measurement. Deep learning could be applied for automatic landmark detection and alignment analysis, with mild to moderate standard errors and favourable correlations with manual measurement. In this study, based on 2210 annotated images of various spinal disease aetiologies, we developed deep learning models capable of automatically locating 45 anatomic landmarks and subsequently generating 18 radiographic parameters on a whole-spine lateral radiograph. In the assessment of model performance, the localisation accuracy and learning speed were the highest for landmarks in the cervical area, followed by those in the lumbosacral, thoracic, and femoral areas. All the predicted radiographic parameters were significantly correlated with ground truth values (all p < 0.001). The human and artificial intelligence comparison revealed that the deep learning model was capable of matching the reliability of doctors for 15/18 of the parameters. The proposed automatic alignment analysis system was able to localise spinal anatomic landmarks with high accuracy and to generate various radiographic parameters with favourable correlations with manual measurements.

scholarly journals AltumAge: A Pan-Tissue DNA-Methylation Epigenetic Clock Based on Deep Learning

2021 ◽  
Author(s):  
Lucas Paulo de Lima ◽  
Louis R Lapierre ◽  
Ritambhara Singh
Keyword(s):  
Dna Methylation ◽  
Deep Learning ◽  
Placental Tissue ◽  
Training Data ◽  
Data Sets ◽  
Epigenetic Clock ◽  
Neural Network Approach ◽  
Cpg Sites ◽  
Age Related ◽  
Model Predictions

Several age predictors based on DNA methylation, dubbed epigenetic clocks, have been created in recent years. Their accuracy and potential for generalization vary widely based on the training data. Here, we gathered 143 publicly available data sets from several human tissues to develop AltumAge, a highly accurate and precise age predictor based on deep learning. Compared to Horvath's 2013 model, AltumAge performs better across both normal and malignant tissues and is more generalizable to new data sets. Interestingly, it can predict gestational week from placental tissue with low error. Lastly, we used deep learning interpretation methods to learn which methylation sites contributed to the final model predictions. We observed that while most important CpG sites are linearly related to age, some highly-interacting CpG sites can influence the relevance of such relationships. We studied the associated genes of these CpG sites and found literary evidence of their involvement in age-related gene regulation. Using chromatin annotations, we observed that the CpG sites with the highest contribution to the model predictions were related to heterochromatin and gene regulatory regions in the genome. We also found age-related KEGG pathways for genes containing these CpG sites. In general, neural networks are better predictors due to their ability to capture complex feature interactions compared to the typically used regularized linear regression. Altogether, our neural network approach provides significant improvement and flexibility to current epigenetic clocks without sacrificing model interpretability.

scholarly journals Automated Ground Truth Generation for Learning-Based Crack Detection on Concrete Surfaces

2021 ◽  
Vol 11 (22) ◽  
pp. 10966
Author(s):  
Hsiang-Chieh Chen ◽  
Zheng-Ting Li
Keyword(s):  
Deep Learning ◽  
Crack Detection ◽  
Ground Truth ◽  
Experimental Results ◽  
Training Data ◽  
Generation Process ◽  
Detection Model ◽  
Ground Truth Generation ◽  
Labeling Approach ◽  
The Cost

This article introduces an automated data-labeling approach for generating crack ground truths (GTs) within concrete images. The main algorithm includes generating first-round GTs, pre-training a deep learning-based model, and generating second-round GTs. On the basis of the generated second-round GTs of the training data, a learning-based crack detection model can be trained in a self-supervised manner. The pre-trained deep learning-based model is effective for crack detection after it is re-trained using the second-round GTs. The main contribution of this study is the proposal of an automated GT generation process for training a crack detection model at the pixel level. Experimental results show that the second-round GTs are similar to manually marked labels. Accordingly, the cost of implementing learning-based methods is reduced significantly because data labeling by humans is not necessitated.

scholarly journals LiDAR Odometry by Deep Learning-based Feature Points with Two-step Pose Estimation

2021 ◽  
Author(s):  
Tianyi Liu ◽  
Yan Wang ◽  
xiaoji niu ◽  
Chang Le ◽  
Tisheng Zhang ◽  
...  
Keyword(s):  
Deep Learning ◽  
Field Experiment ◽  
Pose Estimation ◽  
Ground Truth ◽  
Feature Point ◽  
Training Data ◽  
Lidar Data ◽  
Feature Points ◽  
Experiment Data ◽  
Research And Innovation

KITTI dataset is collected from three types of environments, i.e., country, urban and highway The types of feature point cover a variety of scenes. The KITTI dataset provides 22 sequences of LiDAR data. 11 sequences of them from sequence 00 to sequence 10 are "training" data. The training data are provided with ground truth translation and rotation. In addition, field experiment data is collected by low-resolution LiDAR, VLP-16 in Wuhan Research and Innovation Center.

Sign in / Sign up

Export Citation Format

Share Document