<html>
  <head>
    <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
  </head>
  <body text="#000000" bgcolor="#FFFFFF">
    This is probably the longest-reviewed paper ever submitted by an
    Autonian!<br>
    It feels like we were submitting it sometime a century ago :P<br>
    <br>
    But the ever-elusive GTC paper is finally  out, and I am very proud
    of Mathieu <br>
    who persevered through this challenging experience!<br>
    <br>
    Cheers,<br>
    Artur<br>
    <br>
    PS If this paper is setting the record for the amount of time spent
    in the editor's hands,<br>
    let us keep it this way and avoid challenging it with another
    submission :)<br>
    <br>
    <div class="moz-cite-prefix">On 11/16/2017 2:15 PM, Lujie Chen
      wrote:<br>
    </div>
    <blockquote type="cite"
cite="mid:CAOQ-VV2=9qK+SL2w1v52JeJS1vK9w1dm-eBiXREgKFsz7w5VvA@mail.gmail.com">
      <div dir="ltr">Finally saw it in press! <br>
        <div>
          <div class="gmail_quote">---------- Forwarded message
            ----------<br>
            From: <b class="gmail_sendername">Charles Sutton</b> <span
              dir="ltr"><<a href="mailto:production@jmlr.org"
                moz-do-not-send="true">production@jmlr.org</a>></span><br>
            Date: Thu, Nov 16, 2017 at 1:11 PM<br>
            Subject: [Jmlr-announce] Classification of Time Sequences
            using Graphs of Temporal Constraints<br>
            To: <a href="mailto:jmlr-announce@csail.mit.edu"
              moz-do-not-send="true">jmlr-announce@csail.mit.edu</a><br>
            Cc: <a href="mailto:production@jmlr.org"
              moz-do-not-send="true">production@jmlr.org</a><br>
            <br>
            <br>
            The Journal of Machine Learning Research (<a
              href="http://www.jmlr.org" rel="noreferrer"
              target="_blank" moz-do-not-send="true">www.jmlr.org</a>)
            is pleased to announce the publication of a new paper:<br>
            ------------------------------<wbr>------------------------------<wbr>------------------<br>
            Classification of Time Sequences using Graphs of Temporal
            Constraints<br>
            Mathieu Guillame-Bert, Artur Dubrawski<br>
            JMLR 18(121):1-34, 2017.<br>
            Link: <a href="http://jmlr.org/papers/v18/15-403.html"
              rel="noreferrer" target="_blank" moz-do-not-send="true">http://jmlr.org/papers/v18/15-<wbr>403.html</a><br>
            <br>
            Abstract<br>
            We introduce two algorithms that learn to classify Symbolic
            and Scalar Time Sequences (SSTS); an extension of
            multivariate time series. An SSTS is a set of \emph{events}
            and a set of <em>scalars</em>. An
            <em>event</em> is defined by a symbol and a
            time-stamp. A <em>scalar</em> is defined by a
            symbol and a function mapping a number for each possible
            time stamp of the data. The proposed algorithms rely on
            temporal patterns called Graph of Temporal Constraints
            (GTC). A GTC is a directed graph in which vertices express
            occurrences of specific events, and edges express temporal
            constraints between occurrences of pairs of events.
            Additionally, each vertex of a GTC can be augmented with
            numeric constraints on scalar values. We allow GTCs to be
            cyclic and/or disconnected. The first of the introduced
            algorithms extracts sets of co-dependent GTCs to be used in
            a voting mechanism. The second algorithm builds decision
            forest like representations where each node is a GTC. In
            both algorit!<br>
             hms, extraction of GTCs and model building are interleaved.
            Both algorithms are closely related to each other and they
            exhibit complementary properties including complexity,
            performance, and interpretability. The main novelties of
            this work reside in direct building of the model and
            efficient learning of GTC structures. We explain the
            proposed algorithms and evaluate their performance against a
            diverse collection of 59 benchmark data sets. In these
            experiments, our algorithms come across as highly
            competitive and in most cases closely match or outperform
            state-of-the-art alternatives in terms of the computational
            speed while dominating in terms of the accuracy of
            classification of time sequences.<br>
            <br>
            ------------------------------<wbr>------------------------------<wbr>------------------<br>
            This paper and previous papers are available electronically
            at <a href="http://www.jmlr.org" rel="noreferrer"
              target="_blank" moz-do-not-send="true">http://www.jmlr.org</a>
            in PDF format. The papers of Volumes 1-4 were also published
            in hardcopy by MIT Press; please see <a
              href="http://mitpress.mit.edu/JMLR" rel="noreferrer"
              target="_blank" moz-do-not-send="true">http://mitpress.mit.edu/JMLR</a>
            for details. Volume 5 and subsequent volumes are being
            printed in hardcopy by Microtome Publishing. Please see <a
              href="http://www.mtome.com/Publications/JMLR/jmlr.html"
              rel="noreferrer" target="_blank" moz-do-not-send="true">http://www.mtome.com/<wbr>Publications/JMLR/jmlr.html</a>
            for details and ordering information.<br>
            <br>
            Charles Sutton<br>
            <a href="mailto:production@jmlr.org" moz-do-not-send="true">production@jmlr.org</a><br>
            <br>
            ______________________________<wbr>_________________<br>
            Jmlr-announce mailing list<br>
            <a href="mailto:Jmlr-announce@lists.csail.mit.edu"
              moz-do-not-send="true">Jmlr-announce@lists.csail.mit.<wbr>edu</a><br>
            <a
              href="https://lists.csail.mit.edu/mailman/listinfo/jmlr-announce"
              rel="noreferrer" target="_blank" moz-do-not-send="true">https://lists.csail.mit.edu/<wbr>mailman/listinfo/jmlr-announce</a><br>
          </div>
          <br>
          <br clear="all">
          <br>
          -- <br>
          <div class="gmail_signature" data-smartmail="gmail_signature">
            <div dir="ltr">
              <div>
                <div dir="ltr">
                  <div>
                    <div dir="ltr">
                      <p><font face="arial, helvetica, sans-serif">==================</font></p>
                      <p><span
                          style="font-family:arial,helvetica,sans-serif;font-size:12.8px">Karen
                          (Lujie) Chen</span><br>
                      </p>
                      <font face="arial, helvetica, sans-serif">Ph.D.
                        Candidate in Information Systems, Heinz College</font></div>
                    <div dir="ltr"><span
                        style="font-family:arial,helvetica,sans-serif;font-size:12.8px">PIER
                        (Program of Interdisciplinary Educational
                        Research)</span><font face="arial, helvetica,
                        sans-serif"><br>
                        Member of Auton Lab, Robotics Institute</font></div>
                    <div dir="ltr"><font face="arial, helvetica,
                        sans-serif">Newell-Simon Hall 3124<br>
                        Carnegie Mellon University<br>
                        Pittsburgh, PA 15213</font>
                      <p><br>
                      </p>
                    </div>
                  </div>
                </div>
              </div>
            </div>
          </div>
        </div>
      </div>
    </blockquote>
    <br>
  </body>
</html>