BEGIN:VCALENDAR VERSION:2.0 PRODID:IEEE vTools.Events//EN CALSCALE:GREGORIAN BEGIN:VTIMEZONE TZID:Europe/Rome BEGIN:DAYLIGHT DTSTART:20230326T030000 TZOFFSETFROM:+0100 TZOFFSETTO:+0200 RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=3 TZNAME:CEST END:DAYLIGHT BEGIN:STANDARD DTSTART:20231029T020000 TZOFFSETFROM:+0200 TZOFFSETTO:+0100 RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=10 TZNAME:CET END:STANDARD END:VTIMEZONE BEGIN:VEVENT DTSTAMP:20231106T171740Z UID:00BC26E3-B546-4D40-AEFD-985B6127098A DTSTART;TZID=Europe/Rome:20231026T170000 DTEND;TZID=Europe/Rome:20231026T180000 DESCRIPTION:Machine learning on ultra-low power devices:\napproaches for ac curate and energy efficient inference.\n\nMr. Francesco Daghero\, PhD cand idate DAUIN PoliTO.\n\nMachine Learning (ML) is now central to IoT applica tions\,\noffering low latency and privacy benefits through on-device\nproc essing. However\, constraints on computational and memory\nresources in Io T devices hinder direct ML model deployment. This\npresentation explains m ethods like mixed-precision quantization\,\nbinarization\, and adaptive in ference\, optimizing models for\nembedded systems\, reducing energy usage\ , and maintaining\naccuracy.\n\nQuantum Computing applications tackling no wadays’ problems:\noptimising urban traffic mobility case study\n\nMr An drea Marchesin\, PhD Candidate\, DET\, PoliTO\n\nQuantum Computing (QC) of fers a revolutionary\napproach to solving complex problems beyond traditio nal\nprocessing systems. A novel research work suggests that\napplying QC can improve traffic flow and reduce\ncongestion in modern cities\, bringin g benefits to the\nenvironment and citizens' quality of life.\n\nConstrain ed optimization via controlled multipliers\n\nMr. Simone Pirrera\, PhD Can didate\, DAUIN\, PoliTO\n\nWe propose a control theory-based approach to d evelop novel\noptimization algorithms for constrained non-convex problems. \nSpecifically\, we define a continuous-time system that\, under a\nproper ly defined control action\, converges to a feasible stationary\npoint. We design the control action using PI and feedback\nlinearization\, and we co nduct theoretical analysis to show method's\nconvergence. Finally\, we dem onstrate the method's practical\neffectiveness through numerical examples. \n\nRoom: Maxwell Room\, DET\, Politecnico di Torino\, Corso Duca degli Ab ruzzi\, 24\, Torino\, Piemonte\, Italy LOCATION:Room: Maxwell Room\, DET\, Politecnico di Torino\, Corso Duca degl i Abruzzi\, 24\, Torino\, Piemonte\, Italy ORGANIZER:sb.polito@ieee.org SEQUENCE:11 SUMMARY:PITCHD 2023 EDITION: THE 4th PITCHD URL;VALUE=URI:https://events.vtools.ieee.org/m/381910 X-ALT-DESC:Description: <br /><p><strong>Machine learning on ultra-low powe r devices:</strong><br /><strong>approaches for accurate and energy effici ent inference.</strong></p>\n<p><em>Mr. Francesco Daghero\, PhD candidate DAUIN PoliTO.</em><br /><br />Machine Learning (ML) is now central to IoT applications\,<br />offering low latency and privacy benefits through on-d evice<br />processing. However\, constraints on computational and memory<b r />resources in IoT devices hinder direct ML model deployment. This<br /> presentation explains methods like mixed-precision quantization\,<br />bin arization\, and adaptive inference\, optimizing models for<br />embedded s ystems\, reducing energy usage\, and maintaining<br />accuracy.</p>\n<p>&n bsp\;</p>\n<p><strong>Quantum Computing applications tackling nowadays&rsq uo\; problems:</strong><br /><strong>optimising urban traffic mobility cas e study</strong></p>\n<p><em>Mr Andrea Marchesin\, PhD Candidate\, DET\, P oliTO</em></p>\n<p>Quantum Computing (QC) offers a revolutionary<br />appr oach to solving complex problems beyond traditional<br />processing system s. A novel research work suggests that<br />applying QC can improve traffi c flow and reduce<br />congestion in modern cities\, bringing benefits to the<br />environment and citizens' quality of life.</p>\n<p>&nbsp\;</p>\n< p><strong>Constrained optimization via controlled multipliers</strong></p> \n<p><em>Mr. Simone Pirrera\, PhD Candidate\, DAUIN\, PoliTO</em></p>\n<p> We propose a control theory-based approach to develop novel<br />optimizat ion algorithms for constrained non-convex problems.<br />Specifically\, we define a continuous-time system that\, under a<br />properly defined cont rol action\, converges to a feasible stationary<br />point. We design the control action using PI and feedback<br />linearization\, and we conduct t heoretical analysis to show method's<br />convergence. Finally\, we demons trate the method's practical<br />effectiveness through numerical examples .</p>\n<p>&nbsp\;</p> END:VEVENT END:VCALENDAR