BEGIN:VCALENDAR VERSION:2.0 X-WR-CALNAME:EventsCalendar PRODID:-//hacksw/handcal//NONSGML v1.0//EN CALSCALE:GREGORIAN BEGIN:VTIMEZONE TZID:America/New_York LAST-MODIFIED:20240422T053451Z TZURL:https://www.tzurl.org/zoneinfo-outlook/America/New_York X-LIC-LOCATION:America/New_York BEGIN:DAYLIGHT TZNAME:EDT TZOFFSETFROM:-0500 TZOFFSETTO:-0400 DTSTART:19700308T020000 RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU END:DAYLIGHT BEGIN:STANDARD TZNAME:EST TZOFFSETFROM:-0400 TZOFFSETTO:-0500 DTSTART:19701101T020000 RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU END:STANDARD END:VTIMEZONE BEGIN:VEVENT CATEGORIES:College of Engineering,Thesis/Dissertations DESCRIPTION:Advisor: Dr. Haiping Xu Committee Members: Dr. Joshua Carberry and Dr. Firas Khatib Despite advances in weather forecasting technology, p oor visibility on the open ocean remains a significant safety hazard. Rapi dly changing maritime weather conditions can cause visibility to plummet w ithin minutes, leading to potential accidents and shipping delays. In rece nt years, deep neural networks (DNNs) have demonstrated strong performance in land-based visibility prediction; however, such approaches typically r ely on large historical datasets and provide forecasts for fixed geographi c areas. Oceanic visibility forecasting is more challenging due to the abs ence of stationary weather stations, constantly changing vessel positions, and the highly dynamic nature of marine environments. To address these ch allenges, this thesis introduces self-adaptive deep learning (SADL) models for real-time marine visibility forecasting using time-series data collec ted from onboard sensors and drone-based atmospheric measurements. The pro posed models incorporate a continuous online learning mechanism that updat es model weights in real time, enabling adaptation to both short-term weat her variations and long-term environmental trends. Real-time training is p erformed using small data batches, ensuring computational efficiency while maintaining high data relevance. The SADL models are built upon the Long Short-Term Memory (LSTM) architecture, which is well-suited for capturing the temporal dependencies in visibility forecasting. The models are evalua ted using high-fidelity simulated data derived from real-world datasets pr ovided by NOAA and Copernicus. Case studies, including a realistic simulat ed storm scenario, demonstrate that the proposed approach achieves high ac curacy and robust performance under diverse and extreme conditions. These results highlight the potential of integrating self-adaptive deep learning with real-time sensor data to enhance maritime safety through accurate vi sibility forecasting. All CIS graduate students are encouraged to attend. For further information \nEvent page: https://www.umassd.edu/events/cms/ci s-masters-thesis-defense-by-william-girard.php X-ALT-DESC;FMTTYPE=text/html:

Advisor: Dr. H aiping Xu

Committee Members: Dr. Joshua Carberry and Dr. Firas Khatib

Despite advances in weather forecasting techn ology\, poor visibility on the open ocean remains a significant safety haz ard. Rapidly changing maritime weather conditions can cause visibility to plummet within minutes\, leading to potential accidents and shipping delay s. In recent years\, deep neural networks (DNNs) have demonstrated strong performance in land-based visibility prediction\; however\, such approache s typically rely on large historical datasets and provide forecasts for fi xed geographic areas. Oceanic visibility forecasting is more challenging d ue to the absence of stationary weather stations\, constantly changing ves sel positions\, and the highly dynamic nature of marine environments. To a ddress these challenges\, this thesis introduces self-adaptive deep learni ng (SADL) models for real-time marine visibility forecasting using time-se ries data collected from onboard sensors and drone-based atmospheric measu rements. The proposed models incorporate a continuous online learning mech anism that updates model weights in real time\, enabling adaptation to bot h short-term weather variations and long-term environmental trends. Real-t ime training is performed using small data batches\, ensuring computationa l efficiency while maintaining high data relevance. The SADL models are bu ilt upon the Long Short-Term Memory (LSTM) architecture\, which is well-su ited for capturing the temporal dependencies in visibility forecasting. Th e models are evaluated using high-fidelity simulated data derived from rea l-world datasets provided by NOAA and Copernicus. Case studies\, including a realistic simulated storm scenario\, demonstrate that the proposed appr oach achieves high accuracy and robust performance under diverse and extre me conditions. These results highlight the potential of integrating self-a daptive deep learning with real-time sensor data to enhance maritime safet y through accurate visibility forecasting.

All CIS graduate studen ts are encouraged to attend. For further information

Event page: https://www.umassd.edu/events/cms/cis-masters-thesis-defe nse-by-william-girard.php

DTSTAMP:20260417T025953 DTSTART;TZID=America/New_York:20260403T150000 DTEND;TZID=America/New_York:20260403T160000 LOCATION:SENG 117 SUMMARY;LANGUAGE=en-us:CIS Masters Thesis Defense by William Girard UID:c084f08dc311e64f08a9cb87cae495c1@www.umassd.edu END:VEVENT END:VCALENDAR