<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=utf-8">
</head>
<body text="#000000" bgcolor="#FFFFFF">
<p>Dear all,</p>
<p>On Thursday (18 October) at 2pm in <span class="event-where">54
/ 7035 (7B), we have an S3RI seminar from </span><span
class="event-where"><span class="event-description">Rebecca
Killick (Lancaster University)</span> on </span><span
class="event-where"><span class="event-description">"</span></span><span
class="event-where"><span class="event-description"><span
class="event-description"><span class="event-description">Computationally
Efficient Multivariate Changepoint Detection with Subsets</span></span>".
Details are given below.<br>
</span></span></p>
<p><span class="event-where"><span class="event-description"><span><span
class="event-description">The seminar will also be
available via a live web-cast at</span></span></span></span><br>
<span class="event-description"><a class="moz-txt-link-freetext" href="https://coursec">https://coursec</a><wbr>ast.soton.ac.uk<wbr>/Panopto/Pages/<wbr>Viewer.aspx?id=<wbr>ec46c95b-eb59-4<wbr>da6-b85f-d81b07<wbr>0328c8</span></p>
<p><span class="event-where"><span class="event-description"><span><span
class="event-description">The talk will be followed by tea
and cake in the staff reading room on level 4 of building
54. </span></span></span></span></p>
<p><span class="event-where"><span class="event-description"><span><span
class="event-description">All are welcome!</span></span></span></span></p>
<p><span class="event-where"><span class="event-description"><span>Best
wishes,</span></span></span></p>
<p><span class="event-where"><span class="event-description"><span>Helen</span></span></span></p>
<span class="event-description">Computationally Efficient
Multivariate Changepoint Detection with Subsets
<br>
<br>
Rebecca Killick, Lancaster University
<br>
<br>
Historically much of the research on changepoint analysis has
focused on the univariate setting. Due to the growing number of
high dimensional datasets there is an increasing need for methods
that can detect changepoints in multivariate time series. In this
talk we focus on the problem of detecting changepoints where only
a subset of the variables under observation undergo a change, so
called subset multivariate changepoints. One approach to locating
changepoints is to choose the segmentation that minimises a
penalised cost function via a dynamic program. The work in this
presentation is the first to create a dynamic program specifically
for detecting changes in subset-multivar<wbr>iate time series. The
computational complexity of the dynamic program means it is
infeasible even for medium datasets. Thus we propose a
computationally efficient approximate dynamic program, SPOT. We
demonstrate that SPOT always recovers a better segmentation, in
terms of penalised cost, then other approaches which assume every
variable changes. Furthermore under mild assumptions the
computational cost of SPOT is linear in the number of data points.
In small simulation studies we demonstrate that SPOT provides a
good approximation to exact methods but is feasible for datasets
that contain thousands of variables observed at millions of time
points. Furthermore we demonstrate that our method compares
favourably with other commonly used multivariate changepoint
methods and achieves a substantial improvement in performance when
compared with fully multivariate methods.</span><br>
<br>
<span class="event-where"><span class="event-description"><span><span
class="event-description"><span> <span
class="event-description"> </span><span
class="event-description"><span
class="event-description">For the current schedule of
S3RI seminars, see <a class="moz-txt-link-freetext"
href="https://tinyurl.com/s3riseminar">https://tinyurl.com/s3riseminar</a>
</span></span></span></span></span></span></span><br>
<p><span class="event-where"><span class="event-description"><span><span
class="event-description"></span></span></span></span></p>
</body>
</html>