November 2023 - Gradstudents - mailman.cs.vt.edu

Jie Bu's Ph.D. defense [Feb 9 at 4 pm]
by Anuj Karpatne 02 Dec '24

02 Dec '24

Hello everyone, You are cordially invited to the final defense of PhD student Jie Bu on Thursday, Feb 9 at 4 pm. Zoom Link: https://virginiatech.zoom.us/j/5142223333 Title: Achieving More with Less: Learning Generalizable Neural Networks With Less Labeled Data and Computational Overheads Committee: Dr. Anuj Karpatne (advisor) Dr. Adrian Sandu Dr. Ismini Lourentzou Dr. Naren Ramakrishnan Dr. Anish Arora (from Ohio State University) Abstract: Recent advancements in deep learning have demonstrated its incredible ability to learn generalizable patterns and relationships automatically from data in a number of mainstream applications. However, the generalization power of deep learning methods largely comes at the costs of working with very large datasets and using highly compute-intensive models. Many applications cannot afford these costs needed to ensure generalizability of deep learning models. For instance, obtaining labeled data can be costly in scientific applications, and using large models may not be feasible in resource-constrained environments involving portable devices. This thesis aims to improve efficiency in machine learning by exploring different ways to learn generalizable neural networks that require less labeled data and computational resources. We demonstrate that using physics supervision in scientific problems can reduce the need for labeled data, thereby improving data efficiency without compromising model generalizability. Additionally, we investigate the potential of transfer learning powered by transformers in scientific applications as a promising direction for further improving data efficiency through domain adaptation. On the computational efficiency side, we present two efforts for increasing parameter efficiency of neural networks through novel architectures and structured network pruning. Thanks, -Anuj ------------------- Anuj Karpatne Assistant Professor Computer Science, Virginia Tech https://people.cs.vt.edu/karpatne/

2 6

Muskan Gupta's Thesis exam
by Sang Won Lee 02 May '24

02 May '24

Hi, You are cordially invited to Muskan Gupta's thesis exam. I've included the details below. Sang Won Lee *When/Where* 8/31/2023 (Thursday) 1 PM-3 PM The Gilbert Place Room 3211 *Committee* Sang Won Lee (Chair) Eugenia Rho Scott McCrickard *Title*: Understanding Social Media Users' Perceptions of Trigger and Content Warnings *Abstract*: Exposure to distressing content on social media poses risks to users' mental well-being, leading to the use of trigger warnings (TW) and content warnings (CW). However, there is limited insight into user preferences and the factors affecting these warnings' effectiveness. This study addressed this gap through qualitative interviews with 15 participants. Findings revealed generally positive user perceptions of warnings, which provide protection from distressing content, create safe spaces for sharing sensitive experiences, and avoid such content in public spaces. However, understanding when, what, and how to use warnings varied among users, leading to complex decision-making. Factors influencing warning effectiveness included their separation from content, length, awareness, consistency, context, severity, reasoning, creator's intent, and audience knowledge. Participants stressed the need for enhanced education on warnings and triggers in social media. The study produced design recommendations, including personalized filtering, platform guidelines, specificity, nudging creators, interventions against excessive scrolling, accessibility improvements, and community feedback mechanisms. Participants also suggested post-trigger support measures. The study uncovers challenges in balancing warnings with user engagement, specificity, and personalization while avoiding the creation of filter bubbles. While the study spanned various platforms, future research should focus on specific platforms to address limitations like recruitment bias and broad scope. Future directions include incentivizing warning use, implementing nudges and interventions, and improving post-trigger support, as well as investigating audio triggers and quantitative studies for refined TW/CW design to promote a trauma-informed social media environment.

2 2

Md Hossain Shuvo's Ph.D. defense [1 pm on Dec 5th]
by Debswapna Bhattacharya 30 Nov '23

30 Nov '23

Hello everyone, You are cordially invited to the final defense of Ph.D. student Md Hossain Shuvo on Tuesday, December 5 at 1 pm. Location: 3160A Torgersen Hall Zoom Link: https://virginiatech.zoom.us/j/9500679002 <https://virginiatech.zoom.us/j/9500679002> Title: Machine Learning Methods for Protein Model Quality Estimation Abstract: Protein model quality estimation helps at various stages of the computational prediction of protein three-dimensional (3D) structures, which are essential for numerous biological processes. Existing quality estimation methods for monomeric and multimeric protein models fall short in leveraging the latest technological breakthrough in deep learning that has recently revolutionized protein structure prediction, particularly in harnessing the power of a symmetry-aware network informed by the Protein Language Model (pLM) and the integration of important evolutionary and structural information, limiting their ability to improve quality estimation performance. Moreover, despite the significant progress made by AlphaFold2 in predicting highly accurate protein models, existing methods fall short in the estimation of high-resolution protein model quality and its application in improving the accuracy of predicted protein models. Therefore, the objectives of my research include 1) the development of quality estimation methods for monomeric protein models by learning efficient structural representation with the application of the latest deep learning architectures; 2) the application of our quality estimation methods to improve the structural accuracy of computationally generated protein models; and 3) the development of multimeric protein model quality estimation methods with deep graph learning of protein-protein interface. The benefits of achieving these objectives are two-fold for the broader scientific community. Firstly, it will enable the evaluation of protein models without the presence of experimentally determined protein structures. Secondly, our protein model quality estimation methods can potentially guide the overall structure prediction process and improve the structural accuracy of predicted models. Committee: Dr. Debswapna Bhattacharya (advisor) Dr. Lenwood Heath Dr. Liqing Zhang Dr. Alexey Onufriev Dr. Scott Emrich (from the University of Tennessee - Knoxville) Thank you! Debswapna Bhattacharya Associate Professor, Computer Science Virginia Tech https://people.cs.vt.edu/dbhattacharya/

1 0

Travel & Business Meal Reminder
by Teresa Hall 30 Nov '23

30 Nov '23

Hi everyone! Just a friendly reminder, *all travel* (including virtual locations) should be approved by submitting the attached approval form to me. Keep in mind, I can _prepay flight and registration_ expenses on your behalf once you have an approved travel form on file. If registration fees are the only expense incurred, you may submit the receipt alongside your approval form, or as soon as the document is available. If physical travel has taken place, please use the attached travel reimbursement form and submit to me with original receipts after the travel has concluded. *Please _do not_ pay for meals of a fellow traveler. Each traveler who seeks reimbursement is allotted a daily per diem amount to cover personal meal expenses. To claim meals, you will check the boxes on the reimbursement form of the meals you need reimbursed. All meals provided by the hotel or conference should not be claimed. *All transportation costs between places of lodging or business and places where meals are taken are reimbursed through the incidental expense portion of the M&IE per diem. Please do not include these expenses with your reimbursement. For *business meals*, please use the attached reimbursement form to provide the time, date, location, funding source, attendees, and business purpose. _Original, itemized receipts_ are required for this type of reimbursement. If you have questions regarding either reimbursement process, please feel free to contact me anytime. Take care, Teresa -- *Teresa Hall** **Fiscal Technician* Virginia Tech| Department of Computer Science Torgersen Hall, 1180E (0106) 620 Drillfield Drive Blacksburg, VA 24061 540-231-8454 | teresa(a)vt.edu <mailto:teresa@vt.edu> Office Hours: Monday, Wednesday, &Thursday Telework:Tuesday &Friday __

1 0

CS Seminar Friday, December 1, 2:30PM
by Ribbens, Calvin 30 Nov '23

30 Nov '23

Our final seminar of the semester is tomorrow, December 1, 2:30PM in Torgersen 3100 as usual. Our speaker is Dr. Casey Deccio (BYU) who will speak on "Identity, Impersonation, Intrusion -- and other Internet Measurement Adventures<https://website.cs.vt.edu/research/Seminars/Casey-Deccio.html>". Please join us. Zoom<https://virginiatech.zoom.us/j/89995088537?pwd=NCtqMlNBTEV0Q3QvOENVR2FXRDNa…> should be available if you are not able to participate in person. Cal

1 0

FW: FW: Invited Talk -
by Kinder-Potter, Sharon 28 Nov '23

28 Nov '23

Here are the details of the talk: Dr. Junyuan Hong from UT Austin will be presenting an online guest lecture in our class today. The details are below. Dr. Junyuan Hong from UT Austin Title: Foundation Models Meet Data Privacy: Risks and Countermeasures Abstract: Foundation models have become more and more important for Artificial Intelligence but also pose risks when they are fostered from massive data. In training large models, personal information in data could be encoded and raises concerns in data privacy. Therefore, how to measure the risks and design countermeasures is a crucial topic and gains more attention especially with the popularity of Generative AI. In this lecture, I will present three of our recent efforts towards answering the question. First, we developed a general method to measure the privacy risks in model gradients efficiently and effectively. Second, we presented a new security risk that finetuning can seduce text-to-image models to leak more training data. Third, we novelly made Large Language Models privacy-preserving prompt engineer for close-source ones (e.g., ChatGPT) that were not born to protect data privacy. Bio: Junyuan is a postdoctoral fellow in the Institute for Foundations of Machine Learning (IFML) at UT Austin. He obtained Ph.D. degree from Michigan State University and M.S. and B.S. from University of Science and Technology of China. His research interests lie in trustworthy AI, data privacy and AI for healthcare. He has accomplished a series of work that enhanced the fairness, robustness, security and privacy of federated learning. He is also a core member in ILLIDAN Lab team that won the 3rd place in U.S. PETs prize challenge and was covered by the White House. Recently, his researche is centered on the Trustworthy Generative AI with manageable costs and its application in modeling and treating dementia disease. Time: 11AM ET Zoom link: https://virginiatech.zoom.us/j/86594075180?pwd=NDNJUExiS1RUOWpHV0dsUDFlRWFK… Meeting ID: 865 9407 5180 Passcode: 010981 --

1 0

Recruiting participants for a research study with augmented reality (AR) technology
by Matthew Corbett 27 Nov '23

27 Nov '23

The Synergistic Networking And Intelligence Lab (SNAIL) lab invites you to participate in a study designed to test methods to identify and prevent shoulder surfing using eye gaze and augmented reality. After a brief orientation to AR and AR gestures (approximately 10 minutes), participants will be asked to complete a questionnaire on their comfort with entering sensitive information (e.g. passwords or ATM PINs) in various situations. The questionnaire will contain photographic examples of these situations, and a volunteer will also position themselves to simulate these situations in terms of space and orientation. After the completion of this questionnaire, the participant will be asked to wear a Magic Leap 2 Augmented Reality device and evaluate ShouldAR, a system designed to identify and prevent shoulder surfing attacks. This device measures both user and bystander position and eye gaze in order to alert the device user to potential shoulder surfing. The participants will be asked to simulate entering sensitive information in the same scenarios as the previous questionnaire and give feedback to the evaluator on the application’s performance. Finally, participants will answer the same Likert questions as before, on a digital form, assuming they were wearing the device and application they just tested. This will take 10 minutes. The total time required is expected to be 60 minutes. Participants will be given a $20 Amazon gift card for full participation. The study will take place in Gilbert Place, Room 3216, Third Floor, at 220 Gilbert St, Blacksburg, VA 24060. We are seeking participants who are: - are at least 18 years old - have conversational proficiency in written and spoken English - have normal or corrected to normal vision (*contact lenses are supported, glasses are not*) - are willing to conduct basic AR tasks using a Magic Leap 2 in simulated scenarios, and fill out two surveys in English. Participation will be on a first come first serve basis. For more information, please contact Matthew Corbett at matthewc84(a)vt.edu The project is supervised by Dr. Bo Ji in the Department of Computer Science. This experiment has been approved under IRB 23-1135, as required by the Virginia Tech Institutional Review Board.

1 0

CS Grad Council Meeting -- Monday 11/27 from 12:30 to 1:30 PM in Gilbert 3001
by Wesley Woo 26 Nov '23

26 Nov '23

Hi all, Just a reminder that there will be a CS Grad Council Meeting *tomorrow* (11/27) from *12:30 pm to 1:30 pm in Gilbert 3001* -- we will be ordering a couple pizzas from Dominos for meeting attendees! This meeting will be hybrid (link <https://virginiatech.zoom.us/j/84656150447>) and the last of 2023 :') We will be discussing things including, but not limited to: - Organizing a Quals Q and A - Organizing an end of semester event!! (Likely not to be a barbecue due to general December weather) Hope to see everyone there! Wesley Woo CS Grad Council Secretary

1 0

Additional Opportunity for CS Grad Students!
by Watson, Joan 19 Nov '23

19 Nov '23

Apologies, Grad Students! This notification didn't make it to the CS|Source Opportunities and Announcements email I just distributed, but it is time-sensitive, so I wanted to be sure you received it: Wanted: Graduate Research Assistant An Interdisciplinary team of researchers is seeking a Graduate Research Assistant in Computational Spatial Science for Spring 2024. This role is ideal for candidates who are actively pursuing an advanced degree (M.S. or Ph.D.) at Virginia Tech in a quantitative field and are passionate about leveraging their programming and data science expertise applied to real problems. Please see the attached description for more information and application processes. Note: the priority deadline for applications is Friday, November 24, 2023. Enjoy your break! Joan *************** Joan Monahan Watson, Ph.D. Director of Administrative Operations Department of Computer Science Virginia Tech 1180C Torgersen Hall 620 Drillfield Drive Blacksburg, VA 24061 Phone: 540.231.8874 Email: jmwatson(a)vt.edu<mailto:jmwatson@vt.edu>

1 0

Selecting Your Ideal Mobility Scooter: An Exhaustive Guide
by Morissette, Amie 18 Nov '23

18 Nov '23

For individuals facing mobility challenges, choosing the right mobility scooter is a pivotal decision that can greatly enhance their independence and quality of life. This exhaustive guide is designed to assist you in navigating the diverse range of mobility scooters, ensuring that you find a scooter that meets your specific needs and preferences. Evaluating Personal Mobility Needs The first step in selecting a mobility scooter is to thoroughly understand your unique mobility requirements. Reflect on your daily routine, the kind of terrains you frequently encounter, and any specific physical limitations. If your environment predominantly features flat, even surfaces, a smaller scooter might be your best option. However, for more challenging, uneven terrains, a scooter with a sturdier build and extended battery life would be more appropriate. Types of Mobility Scooters * Travel/Compact Scooters: Ideal for those with active lifestyles or limited storage, these scooters are lightweight and easily dismantled for transport. * 3-Wheel Scooters: These scooters provide enhanced maneuverability and additional legroom, making them suitable for indoor usage and smooth outdoor surfaces. * 4-Wheel Scooters: Offering improved stability, these scooters are well-suited for outdoor environments, especially on rougher terrains. * Heavy-Duty Scooters: Catering to larger individuals, these scooters have higher weight capacities and are built for lasting performance and durability. Features to Consider * Battery Life: An important consideration for those who need to use their scooter for long periods without frequent recharging. * Comfort: Adjustable seats, armrests, and steering mechanisms are important for comfort during extended use. * Weight Capacity: It's essential that the scooter can support your weight comfortably. * Portability: If you often travel, a scooter that can be folded or easily disassembled is beneficial. * Wheel and Suspension Quality: Vital for a smooth ride on varied terrains. * Safety Features: Stability features, lighting, and reflective components are crucial for safe operation. Budget and Insurance Factors The cost of mobility scooters varies, influenced by their features and design. Balancing your needs with your budget is key. Additionally, look into insurance options, as some plans, like Medicare Part B, may cover some of the costs if the scooter is medically necessary. Finding a Reliable Vendor Choosing a reputable vendor is critical. Look for vendors that offer robust customer support, service, and maintenance. A highly recommended option is <http://www.everestmobility.com/> www.everestmobility.com<http://www.everestmobility.com/>, known for its wide selection and expert advice. Testing and User Feedback Test driving various models is advisable to find the best fit in terms of comfort and handling. Also, reading reviews and seeking recommendations from current mobility scooter users can offer invaluable insights. Conclusion Selecting the right mobility scooter is a process that requires a deep understanding of your personal needs, an exploration of the different types available, and a consideration of important features in relation to your budget. For a wide selection and professional guidance, consider visiting resources like <http://www.everestmobility.com/> www.everestmobility.com<http://www.everestmobility.com/>. In conclusion, finding the best mobility scooter involves a tailored approach, comprehensive research, and professional advice. The right scooter can significantly improve your mobility and enhance your daily life.

1 0