Participants will be updated as to current DEIJ initiatives at UConn. Additional, participants will be updated as to current national pushes to dismantle DEIJ in higher education and discuss the implications of this push to dismantle DIEJ. Register - https://fins.uconn.edu/secure_inst/workshops/workshop_view.php?ser=3188 (https://fins.uconn.edu/secure_inst/workshops/workshop_view.php?ser=3188)

Students who may benefit from attending a Let's Talk: Mental Health Office Hours session include:Students who want help connecting to resources but are unsure where to begin Students who are looking for advice on a non-clinical issue Students who are unsure about therapy and are curious about what it is like to talk to a therapist Students who may have concerns about the mental health of a friend and seek advice on how to support their friend If a student is not an imminent risk, and is refusing your support in contacting our office, you may also consider contacting the UConn Student CARE Team (https://studentcareteam.uconn.edu/). This session is held by Ashley Smith, M.A. (https://studenthealth.uconn.edu/person/ashley-smith/)

Interpersonal Group for Graduate Students Graduate Students Interpersonal Groups focuses on promoting emotional wellbeing as you balance academics, relationships, family, and personal responsibilities. Groups offer a supportive confidential space to share your concerns, practice skills and get feedback.To join this group therapy session, please call SHaW at 860-486-4700 (tel:+18604864705) This session is held by Carlos- Gonzalez- Martinez, LCSW (https://studenthealth.uconn.edu/person/carlos-gonzalez-martinez/) For many concerns that students face – like overwhelming stress, anxiety, difficult relationships, depression, academic difficulties, and more – group therapy is the best option for support and healing. Facilitated by Student Health and Wellness (SHaW) counselors, our therapy groups encourage peer support, promote emotional wellbeing, and increase a felt sense of connection. Participants often find that they feel less alone in their struggles, and walk away with newfound support and ideas for coping.

In this session, you will learn to set up and manage your Gradebook to accurately record and calculate student grades in the new Ultra Course View of HuskyCT. We will cover specific settings and features such as the different types of calculations, using categories, and working with the different Gradebook views that are available to instructors. We will also cover how to send a grade column to Student Admin for midterm and final grade submission. 09/24/2024 2:30pm - 3:30pm Register - https://fins.uconn.edu/secure_inst/workshops/workshop_view.php?ser=3166 (https://fins.uconn.edu/secure_inst/workshops/workshop_view.php?ser=3166)

Title: Model Sparsification on Emerging Applications and Technologies Ph.D. Candidate: Shanglin Zhou Major Advisor: Dr. Caiwen Ding Co-Major Advisor: Dr. Krishna Pattipati Associate Advisors: Dr. Cunxi Yu, Dr. Zhijie Shi Date/Time: Tuesday, September 24th, 2024, 3:00pm - 4:00pm Location: Virtual Abstract: Deep neural networks (DNNs) with higher accuracy often lead to larger models, increasing storage and energy demands. Model sparsification can reduce size but risks compromising accuracy. Balancing these factors is challenging, especially as traditional processors struggle with the requirements of low-power, real-time DNNs, highlighting the need for more efficient solutions. In this thesis, we explore model sparsification in emerging applications. We propose an optimization approach using Surrogate Lagrangian Relaxation (SLR) for weight sparsification, streamlining the typical time-consuming three-step pipeline. It enables faster convergence and maintains high accuracy, even during hard-pruning, with rapid recovery in retraining. We explore two primary emerging applications: (1) Energy-harvesting devices that require dynamic power management. We introduce EVE, an AutoML framework using SLR-based sparsification to find optimal multi-models with shared weights, reducing memory use and adapting to changing environments. (2) Diffractive Optical Neural Networks (DONNs) that are fast and energy-efficient but suffer from accuracy degradation due to interpixel interactions. We propose a physics-aware optimization framework for DONNs, incorporating SLR-based sparsification and roughness modeling to smooth phase changes and preserve accuracy. Additionally, we extend our research to multi-task learning (MTL) with DONNs, which traditionally requires manual reconfiguration. We propose LUMEN-PRO, an automated MTL framework that utilizes a flexible DONN backbone. By rotating shared layers instead of storing task-specific ones, LUMEN-PRO reduces memory usage and enhances accuracy, enabling efficient and high-performance DONNs across various tasks.

Title: Model Sparsification on Emerging Applications and TechnologiesAbstract:Deep neural networks (DNNs) with higher accuracy often lead to larger models, increasing storage and energy demands. Model sparsification can reduce size but risks compromising accuracy. Balancing these factors is challenging, especially as traditional processors struggle with the requirements of low-power, real-time DNNs, highlighting the need for more efficient solutions.In this thesis, we explore model sparsification in emerging applications. We propose an optimization approach using Surrogate Lagrangian Relaxation (SLR) for weight sparsification, streamlining the typical time-consuming three-step pipeline. It enables faster convergence and maintains high accuracy, even during hard-pruning, with rapid recovery in retraining.We explore two primary emerging applications: (1) Energy-harvesting devices that require dynamic power management. We introduce EVE, an AutoML framework using SLR-based sparsification to find optimal multi-models with shared weights, reducing memory use and adapting to changing environments. (2) Diffractive Optical Neural Networks (DONNs) that are fast and energy-efficient but suffer from accuracy degradation due to interpixel interactions. We propose a physics-aware optimization framework for DONNs, incorporating SLR-based sparsification and roughness modeling to smooth phase changes and preserve accuracy.Additionally, we extend our research to multi-task learning (MTL) with DONNs, which traditionally requires manual reconfiguration. We propose LUMEN-PRO, an automated MTL framework that utilizes a flexible DONN backbone. By rotating shared layers instead of storing task-specific ones, LUMEN-PRO reduces memory usage and enhances accuracy, enabling efficient and high-performance DONNs across various tasks.