<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=iso-8859-1">
<style type="text/css" style="display:none;"><!-- P {margin-top:0;margin-bottom:0;} --></style>
</head>
<body dir="ltr">
<div id="divtagdefaultwrapper" style="font-size:12pt;color:#000000;font-family:Calibri,Helvetica,sans-serif;" dir="ltr">
<p></p>
<div>Dear Everyone,</div>
<div><br>
Next weeks talk will be given by Alex Brown<br>
We will meet in A5:1003 at 13:15<span aria-label="" class="c-mrkdwn__br" data-stringify-type="paragraph-break"></span><b data-stringify-type="bold">.<br>
<br>
</b><span data-stringify-type="bold">Best,<br>
Haakon (on behalf of the organisers)<br>
</span><b data-stringify-type="bold"><br>
Speaker: Alex Brown</b><br>
<b data-stringify-type="bold">Title: A Fast and Efficient Gaussian Process based Kilonova Light Emulator for Multi-Messenger Analysis</b><br>
Kilonovae are electromagnetic optical and infrared transients generated by r-process nucleosynthesis in the wake of a binary neutron star or neutron star black hole mergers. Following the joint detection of GW170817, GRB170817a, and AT2017gfo, which demonstrated
that kilonvae may be key sources of heavy elements in the universe, interest in kilonovae rapidly expanded. Simulating kilonovae is a complex, time intensive, and computationally expensive process that requires sophisticated handling of diverse areas such
as nuclear physics and general relativistic magnetohydrodynamics. However, the data analysis method used to analyze observed kilonovae require models that can be evaluated extremely quickly. For this reason, emulators/surrogate models are currently critical.
Here I present a state-of-the-art surrogate model developed using Gaussian processes and trained on high fidelity kilonova models from POSSIS. Additionally, I present our first steps in moving away from traditionally likelihood-based inference and towards
more sophisticated, likelihood-free inference methods.</div>
<br>
<p></p>
</div>
</body>
</html>