BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Europe/Stockholm
X-LIC-LOCATION:Europe/Stockholm
BEGIN:DAYLIGHT
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20210916T132452Z
LOCATION:Jean Calvin
DTSTART;TZID=Europe/Stockholm:20210708T113000
DTEND;TZID=Europe/Stockholm:20210708T120000
UID:submissions.pasc-conference.org_PASC21_sess114_msa175@linklings.com
SUMMARY:Reinforcement Learning Optimizer for Earthquakes Simulations using
  Fiber Bundle Models
DESCRIPTION:Minisymposium\n\nReinforcement Learning Optimizer for Earthqua
 kes Simulations using Fiber Bundle Models\n\nMonterrubio-Velasco, Modesto,
  Carrasco Jimenez, de la Puente\n\nRupture of any heterogeneous material i
 s a complex physical process difficult to model deterministic due to the n
 umber of unmeasurable parameters involved and the poorly constrained physi
 cal conditions. The lack of long seismic series, due to our short instrume
 ntal recording time, makes it difficult to observe whole seismic cycles. T
 hus, the predictive potential of these phenomena usually becomes insuffici
 ent. One of the main goals is to explore new approaches able to generate a
 ccurate synthetic time series (physically and statistically) aiming to pro
 duce a better understanding of the earthquake phenomenon. In this sense, a
 n earthquake simulator based on the Fiber Bundle Model (FBM) that produces
  synthetic series fulfilling seismic statistical patterns has been recentl
 y developed, in particular, those series related to the mainshock and the 
 aftershock sequences (Monterrubio-Velasco et al. 2019a, 2019b, 2020). This
  new model has been coined as TREMOL (sTochastic Rupture Earthquake MOdeL)
 . The FBM is a model whose algorithm is based upon the interaction of indi
 vidual elements, with particular charge transfer rules and a probability d
 istribution function to describe the intrinsic properties of its constitue
 nt elements. This model offers many advantages and great adaptability to d
 escribe various rupture phenomena, from the modeling of rupture in microsc
 opic composite materials to large-scale rupture phenomena such as earthqua
 kes. One of the most important features of TREMOL is that it requires a de
 ep parameter tuning that can significantly improve the approximation of th
 e synthetic results with respect to the real ones. The correct parameteriz
 ation of TREMOL generates seismic synthetic catalogs consistent with those
  observed in nature, thus adjusting the most important empirical relations
 hips of seismology. Unfortunately the strong stochastic and discrete natur
 e of the FBM hinders the application of classical optimization techniques 
 based on, for example, continuous gradient descent methods. As a promising
  alternative to these approaches, supervised machine learning (ML) classif
 ication algorithms have been recently used to predict the best parameter v
 alues associated with some preselected classes (Monterrubio et al., 2018, 
 Llácer et al., 2020). Those algorithms demonstrate high performance in sol
 ving this problem for the specific aftershock application, producing a syn
 thetic behavior of earthquakes close enough to the observed one. However, 
 note that this optimization strategy is inherently discrete due to the ML 
 classification, requiring in general costly training and producing less ac
 curate results as long as the number of classes increases. More precisely,
  the explored supervised techniques were applied to analyze three paramete
 rs requiring a large amount of pre-executed simulations to train the ML mo
 dels. In cases where the model complexity increases (i.e. increasing the d
 imensionality by adding more features, classes, and spatial dimensions) th
 e previous approach may be computationally unaffordable for optimizing the
  TREMOL model. Trying to overcome some of the aforementioned drawbacks, in
  this work we explore another alternative strategy to optimize the TREMOL 
 model following an artificial intelligence (AI) based approach. Instead of
  performing a supervised method to learn the best parameter class, the key
  idea is building an artificial agent that learns from its own experience 
 (with no supervision) which is the optimal parameter value that maximizes 
 a given goal function. This AI paradigm is known as reinforcement learning
  (RL), where the agent interacts with its environment by taking actions an
 d evaluating a reward signal. The final goal is to learn a policy that tra
 nsforms a current environment state into an action that potentially return
 s the maximum accumulation of rewards, taking into account all the possibi
 lities. Here, we reformulate the RL paradigm as an optimization problem fo
 r the TREMOL environment and build an artificial agent that deals with con
 tinuous actions as the values of the FBM parameters. The RL algorithms wor
 k naturally in high dimensional spaces and benefit from multiple ways of a
 ddressing high-performance implementations, for instance, the possibility 
 of distributing numerous environment instances for those cases where TREMO
 L requires higher computational costs.\n\nDomain: CS and Math, Emerging Ap
 plications, Climate and Weather, Solid Earth Dynamics
END:VEVENT
END:VCALENDAR