BEGIN:VCALENDAR VERSION:2.0 PRODID:-//pretalx//pretalx.com//adass2023//speaker//YJRWKC BEGIN:VTIMEZONE TZID:US/Arizona BEGIN:STANDARD DTSTART:20221106T000000 TZNAME:MST TZOFFSETFROM:-0700 TZOFFSETTO:-0700 END:STANDARD END:VTIMEZONE BEGIN:VEVENT SUMMARY:A new Deep Learning Model for Gamma-ray bursts’ light curves sim ulation - Luca Castaldini\, Riccardo Falco DTSTART;TZID=US/Arizona:20231106T083000 DTEND;TZID=US/Arizona:20231106T083000 DTSTAMP:20260812T202917Z UID:pretalx-adass2023-9GJPBS@pretalx.com DESCRIPTION:AGILE is a space mission launched in 2007 to study X-ray and g amma-ray astronomy. The AGILE Team is developing new detection algorithms for Gamma-Ray Bursts (GRBs) both with classical and machine learning techn ologies. To train or test these algorithms\, it is necessary to have a lar ge GRB dataset\, but usually\, there are not enough real data available. T his problem is also common for the new generation of high-energy astrophys ics projects (such as COSI and CTA). It therefore becomes essential to hav e a system to simulate GRB data. \nThis work aims to develop a Deep Learni ng-based model for generating synthetic GRBs that closely replicate the pr operties and distribution of real GRBs. The dataset obtained using the tra ined model can then be used to develop detection algorithms both with clas sic techniques and with machine learning. To develop this generative model we have to take into account several complexities. The main ones are the huge different temporal behaviours of such GRBs and the lack of data makin g the training harder.\nWe propose a new method for generating GRBs. This model combines the Generative Adversarial Network (GAN) and Variational Au toencoder (VAE) to produce high-quality and structured generative results while preserving latent space structure. We also developed a loss function tailored to our dataset for the reconstruction task. To create the traini ng dataset we extracted the light curves (LC) of GRBs presented in the Fou rth Fermi-GBM Catalog using only long GRBs. This catalogue contains more t han a decade of observations with a total of 3608 GRBs\, captured through 12 NaI and 2 BGO detectors. In this work\, we considered only the NaI det ectors. We used the LCs detected by multiple sodium detectors and related to the same GRB as independent\, to augment the number of samples. We filt ered the LCs by removing outliers and those with missing values\, ending w ith a total of 5964 LCs. Evaluating the distribution of the GRB duration ( through the t90 parameter)\, we set the length of the time series at 220.\ nWe assessed the model's performance by quantifying the dissimilarity betw een the histograms of count rates in synthetic and real GRBs. Additionally \, we conducted quantitative analysis\, employing statistics of the LCs di stribution in both datasets. The results show that the synthetic LCs are g enerated with very similar properties to the real ones. We are working on a conditional version of our model using physical parameters of the GRBs such as the t90 or the fluence to have a more precise generation. This met hod can be used to generate synthetic LCs for other high-energy astronomy projects such as AGILE\, CTA and COSI. LOCATION:Posters URL:https://pretalx.com/adass2023/talk/9GJPBS/ END:VEVENT END:VCALENDAR