ALBERT

Description: 〈span〉〈div〉ABSTRACT〈/div〉We present the results of the application of the active learning method in developing surrogates as physics‐based earthquake ground‐motion simulators. The surrogates, which map input parameters into output values without demanding intensive computations, are an essential part of any parameter optimization, sensitivity, and uncertainty analysis. Artificial neural networks (ANNs), as an example of surrogates, are very effective in estimating any complicated model. ANNs use a set of training data to learn the mapping process. Training data are a set of input parameters and their corresponding output values. Generating training data requires conducting numerous regional scale ground‐motion simulations. These numerical simulations are computationally challenging. Therefore, a step‐by‐step learning method should be employed to reduce the need for generating unnecessary training data. These methods are called active learning. In this study, we use a pool‐based query by committee (QBC) active learning method with effective initialization approach to study the performance of the models in the training process. We use a dataset that is generated for a moderate earthquake on a regional scale for anelastic attenuation studies with the focus on the estimation of peak ground velocity. The results show that active learning provides better performance in reducing generalization error than does passive learning while the same number of training data is used. Variation of performance with an increasing number of training data is significantly less in an active learning approach which indicates its stable and predictable behavior. This study, although limited to one earthquake and a metric, indicates that in developing surrogates as physics‐based earthquake ground‐motion simulators, application of active learning is an important step in reducing computational demands and generating stable predictors.〈/span〉

Description: 〈span〉〈div〉ABSTRACT〈/div〉We present the results of the application of the active learning method in developing surrogates as physics‐based earthquake ground‐motion simulators. The surrogates, which map input parameters into output values without demanding intensive computations, are an essential part of any parameter optimization, sensitivity, and uncertainty analysis. Artificial neural networks (ANNs), as an example of surrogates, are very effective in estimating any complicated model. ANNs use a set of training data to learn the mapping process. Training data are a set of input parameters and their corresponding output values. Generating training data requires conducting numerous regional scale ground‐motion simulations. These numerical simulations are computationally challenging. Therefore, a step‐by‐step learning method should be employed to reduce the need for generating unnecessary training data. These methods are called active learning. In this study, we use a pool‐based query by committee (QBC) active learning method with effective initialization approach to study the performance of the models in the training process. We use a dataset that is generated for a moderate earthquake on a regional scale for anelastic attenuation studies with the focus on the estimation of peak ground velocity. The results show that active learning provides better performance in reducing generalization error than does passive learning while the same number of training data is used. Variation of performance with an increasing number of training data is significantly less in an active learning approach which indicates its stable and predictable behavior. This study, although limited to one earthquake and a metric, indicates that in developing surrogates as physics‐based earthquake ground‐motion simulators, application of active learning is an important step in reducing computational demands and generating stable predictors.〈/span〉

Description: Significant effort has been devoted over the last two decades to the development of various seismic velocity models for the region of southern California, United States. These models are mostly used in forward wave propagation simulation studies, but also as base models for tomographic and source inversions. Two of these models, the community velocity models CVM-S and CVM-H, are among the most commonly used for this region. This includes two alternative variations to the original models, the recently released CVM-S4.26 which incorporates results from a sequence of tomographic inversions into CVM-S, and the user-controlled option of CVM-H to replace the near-surface profiles with a V S 30 -based geotechnical model. Although either one of these models is regarded as acceptable by the modeling community, it is known that they have differences in their representation of the crustal structure and sedimentary deposits in the region, and thus can lead to different results in forward and inverse problems. In this paper, we evaluate the accuracy of these models when used to predict the ground motion in the greater Los Angeles region by means of an assessment of a collection of simulations of recent events. In total, we consider 30 moderate-magnitude earthquakes (3.5 〈 M w 〈 5.5) between 1998 and 2014, and compare synthetics with data recorded by seismic networks during these events. The simulations are done using a finite-element parallel code, with numerical models that satisfy a maximum frequency of 1 Hz and a minimum shear wave velocity of 200 m s –1 . The comparisons between data and synthetics are ranked quantitatively by means of a goodness-of-fit (GOF) criteria. We analyse the regional distribution of the GOF results for all events and all models, and draw conclusions from the results and how these correlate to the models. We find that, in light of our comparisons, the model CVM-S4.26 consistently yields better results.