arXiv subjects

Oleg Zlydenko

Publications and source records attributed to Oleg Zlydenko.

3 recordsLinked to original sources

arXiv2026

Earthquake magnitudes depend on seismic history, as revealed by a neural network analysis

Earthquake occurrence is notoriously difficult to predict. While some aspects of their spatiotemporal statistics can be relatively well captured by point-process models, very little is known regarding the magnitude of future events, and it is deeply debated whether it is possible to predict the magnitude of an earthquake before it starts. Most operational forecasting models assume that earthquake magnitudes follow a time-independent Gutenberg-Richter (GR) distribution, effectively treating magnitudes as independent of seismic history. We address this fundamental question by demonstrating that standard hypocenter catalogs carry information about future earthquake magnitudes, making them more predictable than previously considered. We present MAGNET (MAGnitude Neural EsTimation model), which uses a multi-encoder neural network architecture with LSTM units to process spatiotemporal patterns in seismic history. By analyzing hypocenter locations, occurrence times, and magnitudes of past events, MAGNET generates probabilistic magnitude forecasts that demonstrate information gains in predicting magnitudes of future events over GR-based models, after controlling for detection artifacts. Our model achieves an information gain of approximately 0.07 bit per earthquake on average over the GR benchmark in Southern California, Japan, and New Zealand catalogs, with this advantage persisting. These results demonstrate that hypocentral earthquake catalogs contain extractable information about future magnitudes, challenging the conventional separability assumption in earthquake forecasting and offering new approaches for seismic hazard assessment.

physics.geo-ph↗

arXiv2020

Incidences between points and curves with almost two degrees of freedom

We study incidences between points and algebraic curves in three dimensions, taken from a family $C$ of curves that have almost two degrees of freedom, meaning that every pair of curves intersect in $O(1)$ points, for any pair of points $p$, $q$, there are only $O(1)$ curves of $C$ that pass through both points, and a pair $p$, $q$ of points admit a curve of $C$ that passes through both of them iff $F(p,q)=0$ for some polynomial $F$. We study two specific instances, one involving unit circles in $R^3$ that pass through some fixed point (so called anchored unit circles), and the other involving tangencies between directed points (points and directions) and circles in the plane; a directed point is tangent to a circle if the point lies on the circle and the direction is the tangent direction. A lifting transformation of Ellenberg et al. maps these tangencies to incidences between points and curves in three dimensions. In both instances the curves in $R^3$ have almost two degrees of freedom. We show that the number of incidences between $m$ points and $n$ anchored unit circles in $R^3$, as well as the number of tangencies between $m$ directed points and $n$ arbitrary circles in the plane, is $O(m^{3/5}n^{3/5}+m+n)$. We derive a similar incidence bound, with a few additional terms, for more general families of curves in $R^3$ with almost two degrees of freedom. The proofs follow standard techniques, based on polynomial partitioning, but face a novel issue involving surfaces that are infinitely ruled by the respective family of curves, as well as surfaces in a dual 3D space that are infinitely ruled by the respective family of suitably defined dual curves. The general bound that we obtain is $O(m^{3/5}n^{3/5}+m+n)$ plus additional terms that depend on how many curves or dual curves can lie on an infinitely-ruled surface.

cs.CG↗

arXiv2019

ML for Flood Forecasting at Scale

Effective riverine flood forecasting at scale is hindered by a multitude of factors, most notably the need to rely on human calibration in current methodology, the limited amount of data for a specific location, and the computational difficulty of building continent/global level models that are sufficiently accurate. Machine learning (ML) is primed to be useful in this scenario: learned models often surpass human experts in complex high-dimensional scenarios, and the framework of transfer or multitask learning is an appealing solution for leveraging local signals to achieve improved global performance. We propose to build on these strengths and develop ML systems for timely and accurate riverine flood prediction.

cs.LG↗