model tuning
Recently Published Documents


TOTAL DOCUMENTS

193
(FIVE YEARS 64)

H-INDEX

18
(FIVE YEARS 5)

2022 ◽  
Vol 6 (GROUP) ◽  
pp. 1-12
Author(s):  
Cathrine Seidelin ◽  
Therese Moreau ◽  
Irina Shklovski ◽  
Naja Holten Møller

As more and more governments adopt algorithms to support bureaucratic decision-making processes, it becomes urgent to address issues of responsible use and accountability. We examine a contested public service algorithm used in Danish job placement for assessing an individual's risk of long-term unemployment. The study takes inspiration from cooperative audits and was carried out in dialogue with the Danish unemployment services agency. Our audit investigated the practical implementation of algorithms. We find (1) a divergence between the formal documentation and the model tuning code, (2) that the algorithmic model relies on subjectivity, namely the variable which focus on the individual's self-assessment of how long it will take before they get a job, (3) that the algorithm uses the variable "origin" to determine its predictions, and (4) that the documentation neglects to consider the implications of using variables indicating personal characteristics when predicting employment outcomes. We discuss the benefits and limitations of cooperative audits in a public sector context. We specifically focus on the importance of collaboration across different public actors when investigating the use of algorithms in the algorithmic society.


2022 ◽  
Vol 27 (1) ◽  
pp. 6
Author(s):  
Mariela Cerrada ◽  
Leonardo Trujillo ◽  
Daniel E. Hernández ◽  
Horacio A. Correa Zevallos ◽  
Jean Carlo Macancela ◽  
...  

Gearboxes are widely used in industrial processes as mechanical power transmission systems. Then, gearbox failures can affect other parts of the system and produce economic loss. The early detection of the possible failure modes and their severity assessment in such devices is an important field of research. Data-driven approaches usually require an exhaustive development of pipelines including models’ parameter optimization and feature selection. This paper takes advantage of the recent Auto Machine Learning (AutoML) tools to propose proper feature and model selection for three failure modes under different severity levels: broken tooth, pitting and crack. The performance of 64 statistical condition indicators (SCI) extracted from vibration signals under the three failure modes were analyzed by two AutoML systems, namely the H2O Driverless AI platform and TPOT, both of which include feature engineering and feature selection mechanisms. In both cases, the systems converged to different types of decision tree methods, with ensembles of XGBoost models preferred by H2O while TPOT generated different types of stacked models. The models produced by both systems achieved very high, and practically equivalent, performances on all problems. Both AutoML systems converged to pipelines that focus on very similar subsets of features across all problems, indicating that several problems in this domain can be solved by a rather small set of 10 common features, with accuracy up to 90%. This latter result is important in the research of useful feature selection for gearbox fault diagnosis.


Author(s):  
Evangelos Alevizos ◽  
Athanasios V Argyriou ◽  
Dimitris Oikonomou ◽  
Dimitrios D Alexakis

Shallow bathymetry inversion algorithms have long been applied in various types of remote sensing imagery with relative success. However, this approach requires that imagery with increased radiometric resolution in the visible spectrum is available. The recent developments in drones and camera sensors allow for testing current inversion techniques on new types of datasets. This study explores the bathymetric mapping capabilities of fused RGB and multispectral imagery, as an alternative to costly hyperspectral sensors. Combining drone-based RGB and multispectral imagery into a single cube dataset, provides the necessary radiometric detail for shallow bathymetry inversion applications. This technique is based on commercial and open-source software and does not require input of reference depth measurements in contrast to other approaches. The robustness of this method was tested on three different coastal sites with contrasting seafloor types. The use of suitable end-member spectra which are representative of the seafloor types of the study area and the sun zenith angle are important parameters in model tuning. The results of this study show good correlation (R2>0.7) and less than half a meter error when they are compared with sonar depth data. Consequently, integration of various drone-based imagery may be applied for producing centimetre resolution bathymetry maps at low cost for small-scale shallow areas.


Author(s):  
Osval Antonio Montesinos López ◽  
Abelardo Montesinos López ◽  
Jose Crossa

AbstractThe overfitting phenomenon happens when a statistical machine learning model learns very well about the noise as well as the signal that is present in the training data. On the other hand, an underfitted phenomenon occurs when only a few predictors are included in the statistical machine learning model that represents the complete structure of the data pattern poorly. This problem also arises when the training data set is too small and thus an underfitted model does a poor job of fitting the training data and unsatisfactorily predicts new data points. This chapter describes the importance of the trade-off between prediction accuracy and model interpretability, as well as the difference between explanatory and predictive modeling: Explanatory modeling minimizes bias, whereas predictive modeling seeks to minimize the combination of bias and estimation variance. We assess the importance and different methods of cross-validation as well as the importance and strategies of tuning that are key to the successful use of some statistical machine learning methods. We explain the most important metrics for evaluating the prediction performance for continuous, binary, categorical, and count response variables.


Author(s):  
Evangelos Alevizos ◽  
Athanasios V Argyriou ◽  
Dimitris Oikonomou ◽  
Dimitrios D Alexakis

Shallow bathymetry inversion algorithms have long been applied in various types of remote sensing imagery with relative success. However, this approach requires that imagery with increased radiometric resolution in the visible spectrum is available. The recent developments in drones and camera sensors allow for testing current inversion techniques on new types of datasets. This study explores the bathymetric mapping capabilities of fused RGB and multispectral imagery, as an alternative to costly hyperspectral sensors. Combining drone-based RGB and multispectral imagery into a single cube dataset, provides the necessary radiometric detail for shallow bathymetry inversion applications. This technique is based on commercial and open-source software and does not require input of reference depth measurements in contrast to other approaches. The robustness of this method was tested on three different coastal sites with contrasting seafloor types. The use of suitable end-member spectra which are representative of the seafloor types of the study area and the sun zenith angle are important parameters in model tuning. The results of this study show good correlation (R2>0.7) and less than half a meter error when they are compared with sonar depth data. Consequently, integration of various drone-based imagery may be applied for producing centimetre resolution bathymetry maps at low cost for small-scale shallow areas.


2021 ◽  
Author(s):  
Dejin Xun ◽  
Deheng Chen ◽  
Yitian Zhou ◽  
Volker M. Lauschke ◽  
Rui Wang ◽  
...  

Deep learning-based cell segmentation is increasingly utilized in cell biology and molecular pathology, due to massive accumulation of diverse large-scale datasets and excellent performance in cell representation. However, the development of specialized algorithms has long been hampered by a paucity of annotated training data, whereas the performance of generalist algorithm was limited without experiment-specific calibration. Here, we present a deep learning-based tool called Scellseg consisted of novel pre-trained network architecture and contrastive fine-tuning strategy. In comparison to four commonly used algorithms, Scellseg outperformed in average precision on three diverse datasets with no need for dataset-specific configuration. Interestingly, we found that eight images are sufficient for model tuning to achieve satisfied performance based on a shot data scale experiment. We also developed a graphical user interface integrated with functions of annotation, fine-tuning and inference, that allows biologists to easily specialize their own segmentation model and analyze data at the single-cell level.


2021 ◽  
Vol 7 ◽  
pp. e822
Author(s):  
Zhisheng Yang ◽  
Jinyong Cheng

In the field of deep learning, the processing of large network models on billions or even tens of billions of nodes and numerous edge types is still flawed, and the accuracy of recommendations is greatly compromised when large network embeddings are applied to recommendation systems. To solve the problem of inaccurate recommendations caused by processing deficiencies in large networks, this paper combines the attributed multiplex heterogeneous network with the attention mechanism that introduces the softsign and sigmoid function characteristics and derives a new framework SSN_GATNE-T (S represents the softsign function, SN represents the attention mechanism introduced by the Softsign function, and GATNE-T represents the transductive embeddings learning for attribute multiple heterogeneous networks). The attributed multiplex heterogeneous network can help obtain more user-item information with more attributes. No matter how many nodes and types are included in the model, our model can handle it well, and the improved attention mechanism can help annotations to obtain more useful information via a combination of the two. This can help to mine more potential information to improve the recommendation effect; in addition, the application of the softsign function in the fully connected layer of the model can better reduce the loss of potential user information, which can be used for accurate recommendation by the model. Using the Adam optimizer to optimize the model can not only make our model converge faster, but it is also very helpful for model tuning. The proposed framework SSN_GATNE-T was tested for two different types of datasets, Amazon and YouTube, using three evaluation indices, ROC-AUC (receiver operating characteristic-area under curve), PR-AUC (precision recall-area under curve) and F1 (F1-score), and found that SSN_GATNE-T improved on all three evaluation indices compared to the mainstream recommendation models currently in existence. This not only demonstrates that the framework can deal well with the shortcomings of obtaining accurate interaction information due to the presence of a large number of nodes and edge types of the embedding of large network models, but also demonstrates the effectiveness of addressing the shortcomings of large networks to improve recommendation performance. In addition, the model is also a good solution to the cold start problem.


Mathematics ◽  
2021 ◽  
Vol 9 (23) ◽  
pp. 3120
Author(s):  
Riccardo Cespi ◽  
Renato Galluzzi ◽  
Ricardo A. Ramirez-Mendoza ◽  
Stefano Di Gennaro

This paper presents an active controller for electric vehicles in which active front steering and torque vectoring are control actions combined to improve the vehicle driving safety. The electric powertrain consists of four independent in–wheel electric motors situated on each corner. The control approach relies on an inverse optimal controller based on a neural network identifier of the vehicle plant. Moreover, to minimize the number of sensors needed for control purposes, the authors present a discrete–time reduced–order state observer for the estimation of vehicle lateral and roll dynamics. The use of a neural network identifier presents some interesting advantages. Notably, unlike standard strategies, the proposed approach avoids the use of tire lateral forces or Pacejka’s tire parameters. In fact, the neural identification provides an input–affine model in which these quantities are absorbed by neural synaptic weights adapted online by an extended Kalman filter. From a practical standpoint, this eliminates the need of additional sensors, model tuning, or estimation stages. In addition, the yaw angle command given by the controller is converted into electric motor torques in order to ensure safe driving conditions. The mathematical models used to describe the electric machines are able to reproduce the dynamic behavior of Elaphe M700 in–wheel electric motors. Finally, quality and performances of the proposed control strategy are discussed in simulation, using a CarSim® full vehicle model running through a double–lane change maneuver.


2021 ◽  
Vol 9 ◽  
Author(s):  
Anders Bryn ◽  
Trine Bekkby ◽  
Eli Rinde ◽  
Hege Gundersen ◽  
Rune Halvorsen

Information about the distribution of a study object (e.g., species or habitat) is essential in face of increasing pressure from land or sea use, and climate change. Distribution models are instrumental for acquiring such information, but also encumbered by uncertainties caused by different sources of error, bias and inaccuracy that need to be dealt with. In this paper we identify the most common sources of uncertainties and link them to different phases in the modeling process. Our aim is to outline the implications of these uncertainties for the reliability of distribution models and to summarize the precautions needed to be taken. We performed a step-by-step assessment of errors, biases and inaccuracies related to the five main steps in a standard distribution modeling process: (1) ecological understanding, assumptions and problem formulation; (2) data collection and preparation; (3) choice of modeling method, model tuning and parameterization; (4) evaluation of models; and, finally, (5) implementation and use. Our synthesis highlights the need to consider the entire distribution modeling process when the reliability and applicability of the models are assessed. A key recommendation is to evaluate the model properly by use of a dataset that is collected independently of the training data. We support initiatives to establish international protocols and open geodatabases for distribution models.


Sign in / Sign up

Export Citation Format

Share Document