Beyond Correlation: A Path-Invariant Measure for Seismogram Similarity
Authors:
Joshua Dickey,
Brett Borghetti,
William Junek,
Richard Martin
Abstract:
Similarity search is a popular technique for seismic signal processing, with template matching, matched filters and subspace detectors being utilized for a wide variety of tasks, including both signal detection and source discrimination. Traditionally, these techniques rely on the cross-correlation function as the basis for measuring similarity. Unfortunately, seismogram correlation is dominated b…
▽ More
Similarity search is a popular technique for seismic signal processing, with template matching, matched filters and subspace detectors being utilized for a wide variety of tasks, including both signal detection and source discrimination. Traditionally, these techniques rely on the cross-correlation function as the basis for measuring similarity. Unfortunately, seismogram correlation is dominated by path effects, essentially requiring a distinct waveform template along each path of interest. To address this limitation, we propose a novel measure of seismogram similarity that is explicitly invariant to path. Using Earthscope's USArray experiment, a path-rich dataset of 207,291 regional seismograms across 8,452 unique events is constructed, and then employed via the batch-hard triplet loss function, to train a deep convolutional neural network which maps raw seismograms to a low dimensional embedding space, where nearness on the space corresponds to nearness of source function, regardless of path or recording instrumentation. This path-agnostic embedding space forms a new representation for seismograms, characterized by robust, source-specific features, which we show to be useful for performing both pairwise event association as well as template-based source discrimination with a single template.
△ Less
Submitted 8 October, 2019; v1 submitted 16 April, 2019;
originally announced April 2019.
URSA: A Neural Network for Unordered Point Clouds Using Constellations
Authors:
Mark B. Skouson,
Brett J. Borghetti,
Robert C. Leishman
Abstract:
This paper describes a neural network layer, named Ursa, that uses a constellation of points to learn classification information from point cloud data. Unlike other machine learning classification problems where the task is to classify an individual high-dimensional observation, in a point-cloud classification problem the goal is to classify a set of d-dimensional observations. Because a point clo…
▽ More
This paper describes a neural network layer, named Ursa, that uses a constellation of points to learn classification information from point cloud data. Unlike other machine learning classification problems where the task is to classify an individual high-dimensional observation, in a point-cloud classification problem the goal is to classify a set of d-dimensional observations. Because a point cloud is a set, there is no ordering to the collection of points in a point-cloud classification problem. Thus, the challenge of classifying point clouds inputs is in building a classifier which is agnostic to the ordering of the observations, yet preserves the d-dimensional information of each point in the set. This research presents Ursa, a new layer type for an artificial neural network which achieves these two properties. Similar to new methods for this task, this architecture works directly on d-dimensional points rather than first converting the points to a d-dimensional volume. The Ursa layer is followed by a series of dense layers to classify 2D and 3D objects from point clouds. Experiments on ModelNet40 and MNIST data show classification results comparable with current methods, while reducing the training parameters by over 50 percent.
△ Less
Submitted 23 October, 2018; v1 submitted 14 August, 2018;
originally announced August 2018.
