Researchers from Tokyo Metropolitan University have applied machine-learning techniques to achieve fast, accurate estimates of local geomagnetic fields using data taken at multiple observation points, potentially allowing detection of changes caused by earthquakes and tsunamis.
A deep neural network (DNN) model was developed and trained using existing data; the result is a fast, efficient method for estimating magnetic fields for unprecedentedly early detection of natural disasters. This is vital for developing effective warning systems that might help reduce casualties and widespread damage.
The devastation caused by earthquakes and tsunamis leaves little doubt that an effective means to predict their incidence is of paramount importance. Certainly, systems already exist for warning people just before the arrival of seismic waves; yet, it is often the case that the S-wave (or secondary wave), that is, the later part of the quake, has already arrived when the warning is given. A faster, more accurate means is sorely required to give local residents time to seek safety and minimize casualties.
It is known that earthquakes and tsunamis are accompanied by localized changes in the geomagnetic field. For earthquakes, it is primarily what is known as a piezo-magnetic effect, where the release of a massive amount of accumulated stress along a fault causes local changes in geomagnetic field; for tsunamis, it is the sudden, vast movement of the sea that causes variations in atmospheric pressure. This in turn affects the ionosphere, subsequently changing the geomagnetic field. Both can be detected by a network of observation points at various locations. The major benefit of such an approach is speed; remembering that electromagnetic waves travel at the speed of light, we can instantaneously detect the incidence of an event by observing changes in geomagnetic field.
However, how can we tell whether the detected field is anomalous or not? The geomagnetic field at various locations is a fluctuating signal; the entire method is predicated on knowing what the “normal” field at a location is.
Thus, Yuta Katori and Assoc. Prof. Kan Okubo from Tokyo Metropolitan University set out to develop a method to take measurements at multiple locations around Japan and create an estimate of the geomagnetic field at different, specific observation points. Specifically, they applied a state-of-the-art machine-learning algorithm known as a Deep Neural Network (DNN), modeled on how neurons are connected inside the human brain. By feeding the algorithm a vast amount of input taken from historical measurements, they let the algorithm create and optimize an extremely complex, multi-layered set of operations that most effectively maps the data to what was actually measured. Using half a million data points taken over 2015, they were able to create a network that can estimate the magnetic field at the observation point with unprecedented accuracy.
Given the relatively low computational cost of DNNs, the system may potentially be paired with a network of high sensitivity detectors to achieve lightning-fast detection of earthquakes and tsunamis, delivering an effective warning system that can minimize damage and save lives.
The Latest on: Deep neural networks
via Google News
The Latest on: Deep neural networks
- New deep-learning approach predicts protein structure from amino acid sequence on April 17, 2019 at 8:05 am
To develop a new approach, AlQuraishi applied so-called end-to-end differentiable deep learning ... school calculus equation—arranged as a neural network, with each component of the network ... […]
- Ikon Science launches RokDoc Deep QI Machine Learning Applications and Ji-Fi Depth Domain providing data driven Geomodelling on April 17, 2019 at 5:16 am
... rapidly expanding portfolio of powerful workflows and solutions for the application of modern Machine Learning and Deep Neural Networks to subsurface problems. Deep QI®, Ikon Science's Machine ... […]
- Bringing Deep Learning for Geospatial Applications to Life on April 17, 2019 at 12:38 am
Deep learning is essentially the latest incarnation of machine learning using Deep Neural Networks. Compared to the traditional machine learning algorithm, it is much faster, more flexible, and able ... […]
- AI Is Better Than People At Tracking Biological Neural Networks on April 16, 2019 at 11:50 am
Our deep learning-based algorithm is fast ... in active neuron detection should provide more information about the neural network and behavioral states, and open the door for accelerated progress ... […]
- Training Deep Neural Networks on a GPU with PyTorch on April 16, 2019 at 6:25 am
This post is the fourth in a series of tutorials on building deep learning models with PyTorch, an open source neural networks library. Check out the full series: In the previous tutorial, we ... […]
- Google Open-Sources GPipe Library for Faster Training of Large Deep-Learning Models on April 16, 2019 at 6:12 am
Google AI is open-sourcing GPipe, a TensorFlow library for accelerating the training of large deep-learning models. Deep-neural-networks (DNN) are the tool of choice for solving many AI tasks, such as ... […]
- BSC Researchers Shrink Floating Point Formats to Accelerate Deep Neural Network Training on April 15, 2019 at 11:01 am
Sometimes calculating solutions as precisely as a computer can wastes more CPU resources than is necessary. A case in point is with deep learning. In early stages of training a Deep Neural Network ... […]
- 'Deep learning' casts wide net for novel 2-D materials on April 10, 2019 at 9:51 am
Rouzbeh Shahsavari, an assistant professor of civil and environmental engineering, and Rice graduate student Prabhas Hundi explored the capabilities of neural networks and multilayer perceptrons ... […]
- A Deep Dive into Deep Learning on April 10, 2019 at 9:02 am
Deep learning by complex neural networks lies behind the applications that are finally bringing artificial intelligence out of the realm of science fiction into reality. Voice recognition allows ... […]
via Bing News