AI Summer School

This event is open to the VALIANT Research Community, which includes Vanderbilt University and its partnering institutions. Individuals involved in research, education, etc. are welcome to register, including from our partner academic communities.

AI Summer School

Join us for an immersive week-long summer school focused on the cutting-edge advancements and practical applications of deep learning and AI. This program is designed for enthusiasts and professionals with a strong foundation in programming, specifically those familiar with Python and modern development environments. Each day is structured to provide a comprehensive blend of theoretical foundations, practical insights, and hands-on experiences, utilizing Python, PyTorch, and VSCode.

Participants will have the opportunity to learn from leading experts, engage in thought-provoking discussions, and explore the latest trends and challenges in the field. The program kicks off with a “Deep Learning Jump Start,” introducing the basics of neural networks and essential optimization techniques. As the week progresses, participants will delve into practical issues with real-world data, including data augmentation, diagnostics, and troubleshooting. Advanced models like transformers and large language models will be explored in depth, along with a dedicated day to visual AI, covering diffusion models and advanced generative techniques. We conclude with critical aspects of AI ethics, including privacy, fairness, transparency, and model robustness. This summer school promises a rich learning experience, combining theoretical knowledge with practical applications, ensuring participants are well-equipped to tackle contemporary challenges in AI.

What: AI Summer School / 2024

Where: Stevenson Hall on Vanderbilt’s Campus.

When: August 12-15, 2024 / 10 – 4 pm

Coffee & Food: Of course!

Cost: Free

On Site Logistics

Final Program PDF

For a full campus map, please see: https://www.vanderbilt.edu/map/

Campus Locations

Campus locations

Detailed View of Room Locations

Room locations

Meet the Speakers

AI Summer Speakers

ACCRE Sponsors GPU Access

ACCRE Sponsors GPU Access

Program at a Glance

August
12^th

Day
1:

Deep
learning jump start

August
13^th

Day
2:

Practical
Issues with real-world data

August
14^th

Day
3:

Integrating
advanced models

August
15^th

Day
4:

Visual
AI Deep Dive

9:00

Coffee
& Conversation – Stevenson 3211 & 3238

Coffee
& Conversation -

Stevenson
3211 & 3238

Coffee
& Conversation – Stevenson 3211 & 3238

Coffee
& Conversation – Stevenson 3211 & 3238

10:00

9:45 am Welcome
Welcome with *Dean Krish Roy*
Introduction with *Provost C. Cybele Raver*

Augmentation,

Tricks
and Dark Magic (Dropout, Batch Norm) - Stevenson 4309

Transformer
Arch. from Zero - Stevenson 4309

Co-Design
- Stevenson 4309

10:30

Deep
Learning 101: Neural Networks Basics - Stevenson 4309

Augmentation,

Tricks,
etc. Cont.

Transformer
Arch. Cont.

Co-Design:
Cont.

11:00

Deep
Learning 101: Cont.

Augmentation,

Tricks,
etc. Cont.

Transformer
Arch. Cont.

Break

11:30

Deep
Learning 101: Cont.

Augmentation,

Tricks,
etc. Cont.

Transformer
Arch. Cont.

Privacy
/ Fairness / Transparency - Stevenson 4309

12:00

Lunch
–

Stevenson 3211 & 3238

Lunch
–

Stevenson 3211 & 3238

Lunch
– Stevenson 3211 & 3238

Lunch
–

Stevenson 3211 & 3238

1:00

Publishing in Deep
Learning Land: Confidence, P-values, and Goats

-
Stevenson 4309

Making
the Ethically Right Choices in AI

- Stevenson 4309

Amplify,
Vanderbilt’s open-source Generative AI application

- Stevenson 4309

Domain
Shift and Image Harmonization - Stevenson 4309

1:30

Privacy
in AI - Stevenson 4309

2:00

Coffee
–

Stevenson
3211 & 3238

Coffee
–

Stevenson
3211 & 3238

Coffee
– Stevenson 3211 & 3238

Coffee
–

Stevenson 3211 & 3238

2:30

Optimizers/Theory
Projects - Stevenson 4309

Diagnostics
and Dashboard - Stevenson 4309

GPTs
and Friends - Stevenson 4309

Decoding
AI’s generative mind: An intro to deep generative models

- Stevenson 4309

3:00

Introduction to ACCRE for Research

Stevenson 4309

Using ACCRE for AI/ML Work

Stevenson 4309

3:30

(working groups)

(working groups)

An
Introduction to Bayesian Modeling: Statistics - Stevenson 4309

4:00

Pull-In
Close - Stevenson 4309

Pull-In
Close - Stevenson 4309

Pull-In
Close - Stevenson 4309

Pull-In
Close - Stevenson 4309

Program Details

Deep Learning 101: Neural Networks Basics

Learning Objectives
- Understand the fundamental concepts of neural networks, including neurons, layers, and activation functions.
- Learn to implement and train a basic neural network using Python and PyTorch.
Brief Biography

Dr. Yuankai Huo is an Assistant Professor of Computer Science at Vanderbilt University and Director of the Biomedical Data Representation and Learning Lab (HRLB Lab) in Nashville, Tennessee, USA. He received his B.S. degree in Electrical Engineering from Nanjing University of Posts and Telecommunications (NJUPT) in 2008, and his master's degree in Electrical Engineering from Southeast University in 2011.
From 2011 to 2014, Dr. Huo worked at Columbia University and the New York State Psychiatric Institute as a staff engineer and research officer. He then earned his second master's degree in Computer Science from Columbia University in 2014. Dr. Huo completed his Ph.D. in Electrical Engineering at Vanderbilt University in 2018.
Following his Ph.D., Dr. Huo served as a Research Assistant Professor at Vanderbilt University and later as a Senior Research Scientist at PAII Labs. Since 2020, he has been a faculty member of the Department of Electrical Engineering and Computer Science, and the Data Science Institute, at Vanderbilt University.
His continuing efforts focus on the translational imaging research that integrates large-scale biomedical image data (microscopy images, digital pathology images, radiology images) with clinical data (genetics, phenotypes, electrical medical records (EMR), biomarkers etc.) for the better health care.
Website

https://github.com/yuankaihuo
https://hrlblab.github.io/

Publishing in Deep Learning Land: Confidence, P-values, and Goats

Learning Objectives
- Understand the significance and application of p-values in deep learning research.
- Learn best practices for publishing deep learning research, including statistical validation and reporting.
Brief Biography

Bennett A. Landman, Ph.D., is the Director of VALIANT (Vanderbilt Lab for Immersive AI) and a Professor of Electrical and Computer Engineering at Vanderbilt University, with roles spanning Computer Science, Biomedical Engineering, Biomedical Informatics, Psychiatry, Neurology, and Radiology. He earned his B.S. (’01) and M.Eng. (’02) in Electrical Engineering and Computer Science from MIT and his Ph.D. in Biomedical Engineering (’08) from Johns Hopkins University.
Dr. Landman’s research focuses on leveraging image-processing technologies for large-scale imaging studies to enhance anatomical understanding and personalize medicine. With funding from the NIH, NSF, Department of Defense, and industry, he has published over 500 peer-reviewed papers. Key areas include volumetric MRI and CT scan analysis, multi-modal AI combining multi-omics, imaging, and health records, and advanced lung screening techniques. He has been an active member of the MICCAI Society Challenge Working Group, co-chaired the SPIE Medical Imaging Image Processing conference (2017-2021), and serves as the Editor-in-Chief of the SPIE Journal of Medical Imaging. He founded the Center for Computational Imaging at Vanderbilt and is the Principal Scientist of ImageVU.
Website

http://my.vanderbilt.edu/masi/

Augmentation, Tricks, and Dark Magic (Dropout, Batch Norm)

Learning Objectives
- Understand the techniques of data augmentation and their importance in improving model generalization.
- Learn advanced regularization methods like Dropout and Batch Normalization to enhance model performance.
Brief Biography

Daniel Moyer's research focuses on machine learning applied to medical imaging. He currently works on tracking and reconstruction projects in fetal MRI, multi-site problems in medical image analysis ("harmonization"), and segmentation problems in intra-vascular ultrasound (IVUS). He also has published work on neuroimaging and diffusion weighted MRI models, for which he received the MICCAI Young Scientist Award in 2016.
Website

https://dcmoyer.github.io/

Making the Ethically Right Choices in AI

Learning Objectives
- Identify common ethical problems in deep learning models and their solutions.
- Learn to recognize and make the ethically right decisions, using real-world scenarios as case studies.
Brief Biography

Dr. Susannah Rose is Core Faculty at Vanderbilt University Medical Center (VUMC) and an Associate Professor in Biomedical Informatics. She also holds positions in the Health Policy department and Center for Health Services Research. Previously at Cleveland Clinic, she had leadership roles such as Director of Research for Bioethics and Associate Chief Experience Officer. She has also been a faculty member at Case Western Reserve University and Harvard University, focusing on bioethics. Dr. Rose earned her Ph.D. in Health Policy from Harvard University in 2010, with notable fellowships from the National Institute of Mental Health and the National Cancer Institute. She holds an MS in Bioethics and an MS in Social Work. Formerly, she was a clinical social worker and researcher at Memorial Sloan-Kettering Cancer Center. She has received numerous teaching awards, published two books, and contributed to high-ranked journals such as JAMA Internal Medicine, NEJM, and The Hastings Center Report. Her research areas include medical technology ethics, patient experience, and AI in healthcare. Her work has been funded by NIH, Harvard University, and The Greenwall Foundation, among others.
Website

https://www.vumc.org/cbmes/person/susannah-l-rose-phd

Optimizers / Theory Projects

Learning Objectives
- Explore different optimization algorithms used in deep learning, such as SGD, Adam, and RMSprop.
- Learn how to apply these optimizers in practical deep learning projects using PyTorch.
Brief Biography

Assistant Professor Soheil Kolouri joined Vanderbilt from HRL Laboratories in Malibu, California, where he was a Principal Machine Learning Scientist focusing on various aspects of deep learning. He received his Ph.D. in biomedical engineering in 2015 from Carnegie Mellon University. Kolouri also was a postdoctoral scholar at CMU focusing on transport-based pattern recognition and image modeling approaches for automated analysis of histopathology, MRI, and fMRI images. He develops practical machine learning and computer vision solutions for challenging problems in biomedical signal and image analysis. Kolouri has more than 40 full-length publications, including 13 journal articles and 25 conference papers at top-tier venues. He has contributed to numerous DARPA proposals in machine learning and has successfully secured a total of nearly $15 million in funding as a PI, including three large DARPA projects. At HRL Laboratories, he mentored Ph.D. students from top universities during their internships. Kolouri’s accomplishments include six patent issued in 2020 and 12 patent applications pending.
Website

https://skolouri.github.io/

Diagnostics and Dashboard

Learning Objectives
- Understand the importance of model diagnostics and performance monitoring.
- Learn to create and interpret dashboards for real-time model evaluation and troubleshooting.
Brief Biography

Yehyun Suh is a dedicated Ph.D. student in the Computer Science program at Vanderbilt University and a member of the Vanderbilt Institute for Surgery and Engineering (VISE). His research focuses on the application of computer vision and deep learning within the medical imaging domain, contributing to the uncovering of hidden information in medical images. His work fosters interdisciplinary collaboration with orthopedic surgeons from Vanderbilt University Medical Center (VUMC) and aids in translating AI advancements into real-world clinical practice.
Yehyun conducts his research in the Vision Information Neuroimaging Engineering (VINE) Laboratory under the guidance of Professor Daniel Moyer. Prior to his graduate studies, Yehyun acquired his undergraduate degree in Information and Communication Engineering at Dongguk University, where he worked under the advisement of Professor Woongsup Kim.
Website

https://yehyunsuh.github.io/

Transformer Architecture from Zero

Learning Objectives
- Understand the principles and components of transformer architecture.
- Understand alternative transformer architectures
- Learn the architectures of full generative AI solutions
Brief Biography

Dr. Jesse Spencer-Smith is the Interim Director and Chief Data Scientist of the Data Science Institute at Vanderbilt University. In addition to being a Computer Science Professor in Practice, Dr. Spencer-Smith leads a team of data scientists collaborating with researchers and industry partners while teaching courses on Artificial Intelligence. With a Ph.D. in Cognitive Psychology and Cognitive Science from Indiana University and a B.S. in Computer Science from the University of Florida, he brings a wealth of expertise to his roles. Previously, as the Director of Data Science at HCA Healthcare, he built the company’s first data science team and facilitated data science implementation throughout the enterprise. Dr. Spencer-Smith also served as an Assistant Professor of Psychology in Quantitative Methods at the University of Illinois Urbana-Champaign and was a Beckman Fellow at the Beckman Institute for Advanced Science and Technology. His diverse background and extensive experience continue to drive advancements in data science and AI, empowering organizations and individuals to harness the power of data-driven insights.
Website

https://github.com/jessespencersmith

GPTs and Friends

Learning Objectives
- Understand the evolution and capabilities of GPT models.
- Learn to apply GPT models for text generation, summarization, and other advanced NLP applications.
Brief Biography

Dr. Vishwesh Nath is an Applied Research Scientist at Nvidia. He works with the Clara DLMED Research Team and his research is focused on medical imaging with sub-domains, including AI-Assisted Annotation (DeepGrow 2D & 3D), Neural Architecture Search and Federated Learning.
Prior to joining Nvidia Dr. Nath pursued his Ph.D. and M.S. degrees in Computer Science from the Electrical Engineering & Computer Science department at Vanderbilt University and a Bachelor's in Electrical and Electronics Engineering from Manipal Institute of Technology. His Ph.D. was focused on Neuroimaging with Diffusion-Weighted Magnetic Resonance Imaging (DW-MRI). He has authored over 20+ publications in the field of medical imaging.
Website

https://scholar.google.com/citations?user=D9RP3YUAAAAJ&hl=en

Domain Shift and Image Harmonization

Learning Objectives
- Understand the challenges and techniques of dealing with domain shift in image data.
- Learn methods for image harmonization to ensure consistency and accuracy across different imaging domains.
Brief Biography

Yihao Liu is a research assistant professor in the Department of Electrical and Computer Engineering at The Medical-image Analysis and Statistical Interpretation (MASI) lab, and an AI fellow at The Vanderbilt Lab for Immersive AI Translation (VALIANT). He obtained his PhD from Johns Hopkins University. He has seven years of experience in medical image analysis. Throughout his career, he has contributed to several projects in image registration, segmentation, synthesis, and surface reconstruction, particularly with optical coherence tomography angiography, magnetic resonance, and ultrasound images.
Website

https://scholar.google.com/citations?user=DlOPRbAAAAAJ&hl=en

Decoding AI’s generative mind: An intro to deep generative models

Learning Objectives
- Understand the principles behind advanced generative models like GANs and VAEs.
- Learn to implement and experiment with these models for creative and practical AI applications.
Brief Biography

Dr. Lianrui Zuo is currently a postdoctoral scholar at the Medical-image Analysis and Statistical Interpretation (MASI) lab and an AI scholar at the VALIANT center of Vanderbilt University. He earned his Ph.D. from Johns Hopkins University. His research interests encompass medical imaging, image harmonization, and deep generative models for medical data. He has published over 40 journal and conference articles, with his work as a first author receiving international recognition, including multiple best paper and poster awards at prestigious conferences. Beyond his research, Dr. Zuo values the power of knowledge sharing, having mentored numerous students and projects. He actively seeks collaborative opportunities that foster learning in the field of medical imaging.
Website

https://scholar.google.com/citations?user=GQGYangAAAAJ&hl=en

Privacy in AI

Learning Objectives
- Understand and articulate key privacy concerns associated with the deployment of AI systems.
- Implement privacy-preserving techniques in AI models.
Brief Biography

Luca Bonomi, PhD, is an Assistant Professor of Biomedical Informatics in the School of Medicine at Vanderbilt University.
His research aims at developing privacy-protecting technologies that provide rigorous privacy protection for biomedical applications. Data privacy research is vital in enabling a sustainable and responsible use of health data. Dr. Bonomi has made significant contributions in several emerging areas, including developing innovative methods for integrating fragmented data and effective approaches for data sharing and predictive modeling.
Dr. Bonomi received a PhD in Computer Science and Informatics from Emory University. He completed his postdoctoral training at the Department of Biomedical Informatics at the University of California San Diego.
Dr. Bonomi is a recipient of an NIH K99/R00 Award “SAFEGENOMES: Strong privacy Assurance for Effective GENOME Sharing”.
Website

https://bonomi-research.github.io/

An Introduction to Bayesian Modeling: Statistics

Learning Objectives
- Understand the basics of Bayesian modeling and its applications in AI.
- Learn to implement Bayesian models and interpret their results for robust statistical analysis.
Brief Biography

Bobby Bodenheimer is a professor of Computer Science at Vanderbilt University, with secondary appointments in Electrical and Computer Engineering, Psychology, and Psychology & Human Development. He received a B.S. and M.S. in Electrical Engineering, and a B.A. in Mathematics from the University of Tennessee. He received his Ph.D. in Electrical Engineering under the direction of John Doyle at Caltech, doing research on parameter-varying control systems. He was a visiting research fellow at the Microsoft graphics research group where he worked on motion capture and controllable human animation, and a postdoctoral fellow under Jessica Hodgins in the GVU at the Georgia Institute of Technology.
At Vanderbilt, his research examines virtual and augmented reality, specifically how people act, perceive, locomote, and navigate in virtual and augmented environments. He runs the Learning in Virtual Environments (LiVE) laboratory. He is the recipient of an NSF CAREER award and his research is funded by NSF, NIH, ONR, and DoD. He has been the conference chair of the ACM APGV and the ACM SCA, the program chair of ACM APGV and IEEE VR, and is a member of the steering committee for ACM SAP. A senior member of the IEEE, he is currently Editor-in-Chief of the ACM Transactions on Applied Perception.
Website

https://eecs.vuse.vanderbilt.edu/People/bobbyb/vita.html

Privacy / Fairness / Transparency

Learning Objectives
- Understand the ethical considerations in AI, focusing on privacy, fairness, and transparency.
- Learn techniques to ensure ethical AI practices in model development and deployment.
Brief Biography

Dr. Aldrich is an Associate Professor in the Department of Medicine, Division of Genetic Medicine and holds additional appointments in the Department of Thoracic Surgery and in the Department of Biomedical Informatics. She conducts cutting-edge research to inform the prevention, diagnosis, and treatment of lung cancer for all populations. Her work encompasses two primary areas of focus: 1) understanding genetic and non-genetic determinants of health contributing to lung cancer risk and outcomes in underrepresented populations, and 2) addressing inequities in lung cancer screening. As a leader in lung cancer screening research, her work has played a pivotal role in shaping key health policy guidelines for lung cancer screening issued by the U.S. Preventive Services Task Force (USPSTF).
Collaborative scientific efforts are underway with national colleagues to investigate the genetics of immunotherapy treatment among lung cancer patients, to assess the impact of structural racism on lung cancer inequities, to identify barriers and facilitators to the recruitment of underrepresented populations to lung cancer clinical trials and to construct residential histories to investigate area-level risk factors among populations experiencing disparities. Additional ongoing research is investigating the population genetics of admixed African ancestry populations. Dr. Aldrich also leads a lung cancer focused community advisory board to spearhead community-engaged dissemination strategies addressing inequities in the early detection of lung cancer.
Website

https://medsites.vumc.org/aldrichlab

Amplify, Vanderbilt’s open-source Generative AI application

Learning Objectives
- Exploring Amplify’s Architecture and Features
- Building Custom Assistants within Amplify
Brief Biography

Allen Karns is an IT professional with 25 years of experience in infrastructure, system administration, and cloud architecture. Most recently, he joined the Chancellor’s office team to develop Vanderbilt’s Open-Source generative AI application, Amplify.
He excels in cloud architecture, infrastructure design, and deployments. He is recognized for his ability to engage stakeholders, lead and mentor teams, and enjoys collaborating on innovative projects. His technical skills span Cloud Architectures in AWS and Azure, with expertise in Infrastructure as Code, security frameworks, automation, and cost optimization. He holds certifications including AWS Certified Solutions Architect Professional, Azure Administrator, and ITIL 4 Framework.
Allen's career is marked by recognition for innovation and customer satisfaction, including the 2019 Vanderbilt Pacesetter Award and the Kudos Awards for 2019 and 2021.
Excited about the future of technology, Allen continually seeks to contribute to projects that push the boundaries of what is possible.
Website

https://www.amplifygenai.org/

Introduction to ACCRE Resources for Research (Tuesday) & Using ACCRE for AI/ML Work

Learning Objectives
- Get onboarded to ACCRE and its resources for running AI/ML workflows
- Receive a brief introduction to the cluster
- Introduction to UNIX and SLUM
Brief Biography

Dr. Brandon Soubasis is a Senior Research Scientist at Vanderbilt University’s ACCRE and the Center for Applied AI in Protein Dynamics. Brandon holds a Ph.D.in Physics from Vanderbilt University, where his research focused on particle physics for the CMS collaboration. His expertise lies in using AI, deep learning, and high-performance computing, to the advancement of AI applications in scientific research.
Website

https://www.vanderbilt.edu/accre/

Sponsors

VALIANT, ADVANCE, ACCRE, CS, DSI, ECE, LIVE, VKC, VINSE, VISE& Jean and Alexander Heard Libraries are co-hosting the AI Summer School / 2024 Release.