BEGIN:VCALENDAR VERSION:2.0 PRODID:IEEE vTools.Events//EN CALSCALE:GREGORIAN BEGIN:VTIMEZONE TZID:America/New_York BEGIN:DAYLIGHT DTSTART:20250309T030000 TZOFFSETFROM:-0500 TZOFFSETTO:-0400 RRULE:FREQ=YEARLY;BYDAY=2SU;BYMONTH=3 TZNAME:EDT END:DAYLIGHT BEGIN:STANDARD DTSTART:20251102T010000 TZOFFSETFROM:-0400 TZOFFSETTO:-0500 RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=11 TZNAME:EST END:STANDARD END:VTIMEZONE BEGIN:VEVENT DTSTAMP:20250509T124610Z UID:E620A890-2FE9-40A9-B09C-1641A9B062B7 DTSTART;TZID=America/New_York:20250507T183000 DTEND;TZID=America/New_York:20250507T193000 DESCRIPTION:Deep neural networks make it possible to learn key characterist ics of data without having to assume mathematically tractable models. This in turn results in an ability to compress data in a model-free way. One o f the promising areas for the application of deep learning to the physical layer of communication networks is compression. Hundred-fold compression with a small loss of the CSI in massive MIMO systems has been shown to be both feasible and necessary. However\, model-free and data driven compress ion comes with a downside: the encoding and decoding models need to be tra ined on a large set of CSI arrays indicative of a wide spectrum of propaga tion and environmental conditions. As a result\, in the early stages of th e deployment of deep CSI compression models\, it would be necessary to det ect if and when users’ channels have drifted significantly away from the distribution of the CSI data on which the deep compression model was trai ned. In this paper\, we present both 1) a technique for detecting harmful channel drift and 2) a lightweight scheme for fine-tuning the deep compres sion models to adjust to such shifts. Using public-domain synthetic channe l data as well as 3GPP-compliant simulated data\, we demonstrate the pract icality of our proposed deep compression and detection framework. We close with recommendations for a viable implementation of the proposed drift de tection by the standards bodies.\n\nCo-sponsored by: New Jersey Coast Sect ion Chapters\, ComSoc Chapter\, COM19\, Jt Chp\,ED15/MTT17/PHO36\, Jt. Cha pter\,SP01/CAS04 and Jt. Chapter\,IM09/C16\, and North Jersey Section Chap ter\,COM 19\n\nSpeaker(s): Dr. Kursat Metsav\, \n\nAgenda: \n06:30 p.m. - Introduction of Dr. Kursat Metsav\n\n06:35 p.m. - Presentation by Dr. Kurs at Metsav\n\n07:10 p.m. - Question & Answer session\n\nVirtual: https://ev ents.vtools.ieee.org/m/476641 LOCATION:Virtual: https://events.vtools.ieee.org/m/476641 ORGANIZER:patluri@att.com SEQUENCE:36 SUMMARY:Data-Driven CSI Compression for MIMO Systems and Detecting Feedback Drift URL;VALUE=URI:https://events.vtools.ieee.org/m/476641 X-ALT-DESC:Description: <br /><p><span style="font-size: 11.0pt\; font-fami ly: 'Aptos'\,sans-serif\; mso-fareast-font-family: Aptos\; mso-fareast-the me-font: minor-latin\; mso-bidi-font-family: Aptos\; mso-ligatures: standa rdcontextual\; mso-ansi-language: EN-US\; mso-fareast-language: EN-US\; ms o-bidi-language: AR-SA\;">Deep neural networks make it possible to learn k ey characteristics of data without having to assume mathematically tractab le models. This in turn results in an ability to compress data in a model- free way. One of the promising areas for the application of deep learning to the physical layer of communication networks is compression. Hundred-fo ld compression with a small loss of the CSI in massive MIMO systems has be en shown to be both feasible and necessary. However\, model-free and data driven compression comes with a downside: the encoding and decoding models need to be trained on a large set of CSI arrays indicative of a wide spec trum of propagation and environmental conditions. As a result\, in the ear ly stages of the deployment of deep CSI compression models\, it would be n ecessary to detect if and when users&rsquo\; channels have drifted signifi cantly away from the distribution of the CSI data on which the deep compre ssion model was trained. In this paper\, we present both 1) a technique fo r detecting harmful channel drift and 2) a lightweight scheme for fine-tun ing the deep compression models to adjust to such shifts. Using public-dom ain synthetic channel data as well as 3GPP-compliant simulated data\, we d emonstrate the practicality of our proposed deep compression and detection framework. We close with recommendations for a viable implementation of t he proposed drift detection by the standards bodies.</span></p><br /><br / >Agenda: <br /><p>06:30 p.m. - Introduction of Dr. Kursat Metsav</p>\n<p>0 6:35 p.m. - Presentation by Dr. Kursat Metsav</p>\n<p>07:10 p.m. - Questio n &amp\; Answer session</p> END:VEVENT END:VCALENDAR