BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Northeastern University College of Engineering - ECPv6.15.20//NONSGML v1.0//EN
CALSCALE:GREGORIAN
METHOD:PUBLISH
X-ORIGINAL-URL:https://coe.northeastern.edu
X-WR-CALDESC:Events for Northeastern University College of Engineering
REFRESH-INTERVAL;VALUE=DURATION:PT1H
X-Robots-Tag:noindex
X-PUBLISHED-TTL:PT1H
BEGIN:VTIMEZONE
TZID:America/New_York
BEGIN:DAYLIGHT
TZOFFSETFROM:-0500
TZOFFSETTO:-0400
TZNAME:EDT
DTSTART:20230312T070000
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0400
TZOFFSETTO:-0500
TZNAME:EST
DTSTART:20231105T060000
END:STANDARD
BEGIN:DAYLIGHT
TZOFFSETFROM:-0500
TZOFFSETTO:-0400
TZNAME:EDT
DTSTART:20240310T070000
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0400
TZOFFSETTO:-0500
TZNAME:EST
DTSTART:20241103T060000
END:STANDARD
BEGIN:DAYLIGHT
TZOFFSETFROM:-0500
TZOFFSETTO:-0400
TZNAME:EDT
DTSTART:20250309T070000
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0400
TZOFFSETTO:-0500
TZNAME:EST
DTSTART:20251102T060000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20240816T103000
DTEND;TZID=America/New_York:20240816T113000
DTSTAMP:20260516T095935
CREATED:20240820T181017Z
LAST-MODIFIED:20240820T181017Z
UID:45115-1723804200-1723807800@coe.northeastern.edu
SUMMARY:Faruk Volkan Mutlu PhD Dissertation  defense
DESCRIPTION:Name:\nFaruk Volkan Mutlu \nTitle:\nCost-aware Joint Caching and Forwarding in Networks with Diverse Cache Resources \nDate:\n8/16/2024 \nTime:\n10:30:00 AM \nCommittee Members:\nProf. Edmund Yeh (Advisor)\nProf. Stratis Ioannidis\nProf. Elif Uysal \nAbstract:\nThe rapid growth of data-intensive applications is testing the limitations of today’s data distribution networks. Caching is a crucial tool for high performance in such networks\, and a core principle in emerging paradigms like information-centric networking (ICN). In this dissertation\, motivated by the needs of a landmark initiative addressing the networking challenges faced by large-scale scientific research programs\, we focus on the key issue of expanding cache capacities in a cost-effective manner. While DRAM is still the standard cache device today due to its high transfer rates\, its capacity is very limited and subject to contention by other networking functions. Large DRAM modules are also expensive\, making wide area networks with cache-enabled routers costly to deploy. On the other hand\, technologies like flash storage offer larger capacities at lower costs; observing recent advancements in this domain\, we expect devices like NVMe SSDs to feature as additional cache tiers in networks supporting data-intensive applications. However\, the slower transfer rates of such devices and the added operational costs they introduce pose some challenges. \nThis dissertation primarily focuses on the open problem of developing cost-aware caching policies that can effectively manage multiple types of cache available to routers. We begin by introducing an object-level multi-tiered caching model that incorporates the diverse characteristics of cache devices\, namely their transfer rates and utilization costs. We then integrate this model with an established optimization framework to develop a joint caching and forwarding policy that uses caching resources available in the network intelligently to improve performance\, minimize costs of cache utilization and avoid congestion. To highlight the advantages of our approach against adapted baselines\, we conduct an exhaustive experimental evaluation of this policy under a large variety of simulation settings; we also provide a discussion of the event-driven object-level ICN simulator we built to facilitate this evaluation and support future research. Lastly\, we also present a variation on our strategy that attempts to improve its efficiency under certain conditions by introducing new control variables into the aforementioned optimization framework. \nThis dissertation also discusses our work on the effective use of transmission and cache resources in wireless networks. While this work investigates a different context than that outlined above\, it serves to complement our primary scope with a perspective on how caching can be leveraged in settings where resource constraints have a different type of interplay with network performance. Specifically\, in the context of multi-hop wireless networks with arbitrary topologies and interfering transmissions\, we study the open problem of delay minimization via caching subject to transmission power limitations. We cast this scenario as an optimization problem and highlight its analytical properties. We identify the challenges in finding a global optimum to the problem under general conditions\, and propose an algorithm that converges to a local optimum. We conclude this discussion with numerical results that demonstrate the effectiveness of our approach. \n 
URL:https://coe.northeastern.edu/event/faruk-volkan-mutlu-phd-dissertation-defense/
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20240816T110000
DTEND;TZID=America/New_York:20240816T123000
DTSTAMP:20260516T095935
CREATED:20240731T141728Z
LAST-MODIFIED:20240731T141728Z
UID:44750-1723806000-1723811400@coe.northeastern.edu
SUMMARY:CommLab Drop-In Writing Hours
DESCRIPTION:Graduate students\, are you looking for a place for focused research writing time?  Join the CommLab drop-in writing hours any Friday from 11 am-12:30 pm ET.  Drop in any Friday and stay for a short time or the whole hour and a half.  CommLab Fellows will be available to provide feedback on your writing.  We will be meeting in 13 International Village.
URL:https://coe.northeastern.edu/event/commlab-drop-in-writing-hours-2/2024-08-16/
LOCATION:13 International Village\, 360 Huntington Ave\, 13 INV\, Boston\, MA\, 02115\, United States
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20240816T110000
DTEND;TZID=America/New_York:20240816T130000
DTSTAMP:20260516T095935
CREATED:20240820T175800Z
LAST-MODIFIED:20240820T175800Z
UID:45125-1723806000-1723813200@coe.northeastern.edu
SUMMARY:Yanyu Li PhD Dissertation Defense
DESCRIPTION:Name:\nYanyu Li \nTitle:\nAccelerating Large Scale Generative AI: a Comprehensive Study \nDate:\n8/16/2024 \nTime:\n11:00:00 AM \nCommittee Members:\nProf. Yanzhi Wang (Advisor)\nProf. David Kaeli\nProf. Kaushik Chowdhury \nAbstract:\nWe have witnessed the great success of deep learning in various domains\, such as the emerging large language models (LLMs) and Artificial General Intelligence (AGI)\, diffusion models for image and video generation\, and classic vision tasks including classification\, segmentation\, detection\, etc. Built with linear\, convolution\, and attention blocks\, Deep Neural Networks (DNNs) play a vital role in the performance revolution. However\, powerful DNNs often call for tremendous computation and storage size\, which hinders their wide adoption. For instance\, LLMs and diffusion models generally have billions of parameters and hundreds of GMACs\, which is prohibitive for edge deployment. As a result\, Efficient AI has become a hot research area. In this work\, with algorithm optimizations and co-designs with hardware platform\, we pursue the appealing features of edge or user-end AI\, where we cut down energy consumption\, shorten response latency\, shrink model storage size\, eliminate the need for cloud server access and protect user privacy. Firstly\, we systematically investigate quantization\, pruning\, and architecture search techniques for efficient vision backbones. We do a comprehensive study on quantization number system and precision\, and propose a novel mix-scheme mix-precision quantization technique to maximize hardware utilization and minimize performance loss. Regarding network pruning\, we propose a novel indicator-based approach\, named Pruning-as-Search\, that is fully differentiable and automatically decides pruning policies\, outperforming human tuning methods in terms of performance and efficiency. Further\, we address the long-existing issue of rigid network width design\, proposing a family of flexible-width pruned networks with minimal per-layer redundancy. As for architecture search\, we formulate a joint optimization objective of both size and latency\, releasing a series of efficient Vision Transformers\, named EfficientFormer (V1 and V2)\, to serve as strong vision backbones with MobileNet-level size and millisecond-level latency on mobile phones. \nSecondly\, we make dedicated optimizations for large-scale generative tasks\, i.e.\, Stable Diffusion (SD) for text-to-image generation\, which serves as pioneer work to enable their mobile deployment. With the proposed efficient architecture design and novel step distillation\, we shrink the generation latency of SD by a magnitude\, from more than 1 minute to generate a 512$\times$512 image to 1~2 seconds\, while preserving the stunning generative quality. We extend our work to the even more challenging video generation task\, enabling 2-bit inference and single step adversarial distillation to speedup video diffusion models by a magnitude.
URL:https://coe.northeastern.edu/event/yanyu-li-phd-dissertation-defense/
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20240816T140000
DTEND;TZID=America/New_York:20240816T160000
DTSTAMP:20260516T095935
CREATED:20240820T182221Z
LAST-MODIFIED:20240820T182221Z
UID:45097-1723816800-1723824000@coe.northeastern.edu
SUMMARY:Shuo Jiang PhD Dissertation Defense
DESCRIPTION:Name:\nShuo Jiang \nTitle:\nTactile Intelligence in Robotics \nDate:\n8/16/2024 \nTime:\n2:00:00 PM \nLocation:\nEXP-701A \nCommittee Members:\nProf. Lawson Wong (Advisor)\nProf. Robert Platt\nProf. Alireza Ramezani\nProf. Taskin Padir \nAbstract:\nIn recent years\, the evolution of robot electronic skin technology has introduced a novel avenue for robots to perceive their external environment and internal state. In contrast to conventional visual perception methods\, tactile perception enables the discernment of additional physical properties of objects\, such as friction and mass distribution\, or even observes contact with higher resolution. Importantly\, tactile perception is resilient to challenges posed by inadequate illumination or environmental occlusion. However\, it presents inherent challenges\, including a limited sensing range\, compulsory physical interaction with the environment\, and intricate coupling with robot control\, rendering data collection and utilization challenging. Addressing these challenges and devising effective\, efficient\, and interpretable methods for processing tactile signals have emerged as pivotal issues in robot tactile perception. \nWith the development of artificial intelligence technology\, we are now able to interpret tactile information from a new perspective beyond traditional sensor technology and signal processing methods\, thereby expanding a wider range of robotic applications. With our continuous efforts over the past few years\, we have comprehensively addressed the following challenges in enhancing robot tactile perception through the application of advanced artificial intelligence and control methods: enabling robots to explore object shapes through tactile feedback; developing tactile-based safety mechanisms for human-robot collaboration; enhancing the locomotion adaptability of snake robots on irregular terrains through tactile perception; utilizing whole-body exteroceptors and proprioceptors for accurate body schema estimation; and implementing tactile gesture recognition in human-robot interactions. At the same time\, we developed a modular full-body electronic skin system for robots and its accompanying software\, which can accurately detect forces applied to the robot’s entire body and perform high-speed tracking of the real-time kinematics of the robot’s sensor array. \nIn conclusion\, this dissertation explores how robot tactile perception can accomplish complex tasks in various scenarios or achieve performance improvements in traditional tasks through the integration of sensor technology\, machine learning\, control theory\, and robotics. Through extensive theoretical and experimental analysis\, we have demonstrated the critical role of tactile perception in embodied intelligence for robots and established a fundamental knowledge framework for future academic research in this field.
URL:https://coe.northeastern.edu/event/shuo-jiang-phd-dissertation-defense/
END:VEVENT
END:VCALENDAR